Forthcoming and Online First Articles

International Journal of Computational Systems Engineering

International Journal of Computational Systems Engineering (IJCSysE)

Forthcoming articles have been peer-reviewed and accepted for publication but are pending final changes, are not yet published and may not appear here in their final order of publication until they are assigned to issues. Therefore, the content conforms to our standards but the presentation (e.g. typesetting and proof-reading) is not necessarily up to the Inderscience standard. Additionally, titles, authors, abstracts and keywords may change before publication. Articles will not be published until the final proofs are validated by their authors.

Forthcoming articles must be purchased for the purposes of research, teaching and private study only. These articles can be cited using the expression "in press". For example: Smith, J. (in press). Article Title. Journal Title.

Articles marked with this shopping trolley icon are available for purchase - click on the icon to send an email request to purchase.

Online First articles are published online here, before they appear in a journal issue. Online First articles are fully citeable, complete with a DOI. They can be cited, read, and downloaded. Online First articles are published as Open Access (OA) articles to make the latest research available as early as possible.

Open AccessArticles marked with this Open Access icon are Online First articles. They are freely available and openly accessible to all without any restriction except the ones stated in their respective CC licenses.

Register for our alerting service, which notifies you by email when new issues are published online.

International Journal of Computational Systems Engineering (89 papers in press)

Regular Issues

  • Research on English teaching curriculum model based on MMSP algorithm   Order a copy of this article
    by Qun Yang  
    Abstract: In order to analyse and extract teaching performance data, this study proposes a multi-dimensional and multi-layer association rule mining multi-dimensional sequential patterns (MMSP) algorithm for mining association rules between college English course learning and other courses. In the design of the MMSP algorithm, the Diffset strategy is used to mine subsets in the most orderly manner, the PrefixSpan algorithm is used to generate frequent sequences, and the mining of multidimensional frequent sequences is transformed into the mining of frequent fundamental vectors. The results indicate that single semester English course learning is influenced by professional and theoretical courses, and the learning situation of multi semester English courses has a certain degree of stability, which is related to the learning of public courses. The unique contribution of this study lies in considering the characteristics of teaching performance data and designing frequent multidimensional sequences to include different levels of hierarchical granularity.
    Keywords: association rule mining; multi-dimensional and multi-layer; educational data; English teaching.
    DOI: 10.1504/IJCSYSE.2025.10059044
  • A network teaching model of College English Viewing, listening and speaking from the perspective of mobile technology   Order a copy of this article
    by Siyi Chen, Xinli Ke, Xiaohong Zhang 
    Abstract: The development of technology has provided more possibilities for educational models. To address the problem that students’ learning cannot be fully taken into account in traditional classrooms, the study designs an English audio-visual network teaching model based on modern mobile technology. The algorithmic model is based on the Adaptive Mutation Genetic Model-Back Propagation Neural Network (AGA-BP) to intelligently estimate the English teaching quality. The model was then validated by means of experiments. It can be showed from this that the mean square error value of the model was 3.4994e*10-12 and the fitness value was 1.25. Compared with the traditional neural network model, the accuracy of the improved model is increased by 16.31% and 6.62% respectively. The experiments show that the designed models have excellent evaluation performance and can be practically applied in specific teaching.
    Keywords: mobile technology; web-based teaching; quality evaluation; AGA-BP neural network.
    DOI: 10.1504/IJCSYSE.2024.10058062
  • Construction of a web-based mathematical model for blended teaching of English in colleges and universities   Order a copy of this article
    by Xuemei Wang  
    Abstract: In the context of continuous higher education reform, blended teaching is gradually applied in the teaching of various majors. However, the traditional teaching quality evaluation system is not suitable for the current college English blended teaching mode. Therefore, the research uses BP neural network and GA algorithm to construct two new hybrid teaching quality evaluation methods based on establishing a hybrid teaching quality evaluation system for college English, and verifies them using experiments. The test results showed that the average score of the BP simulation experiment was 88.14, and the average deviation was 5.16. Most of the prediction errors of BP simulation were below 10 points. While the average score of GA-BP simulation experiments is 86.30 with a relative error of 0.04. In the comparison of four different algorithm models, the scores of the genetic algorithm and BSA algorithm remain between 73 and 105, both of which have a lower score than the BP algorithm, but also have an extreme score. Taken together, the GA-BP neural network-based English blended teaching quality assessment model has a lower error and higher assessment accuracy, which is more scientific for the assessment of the actual university English blended teaching quality.
    Keywords: university English; mixed learning; evaluation model; index system.
    DOI: 10.1504/IJCSYSE.2025.10059045
  • Practical Analysis of Digital Technology of Electronic Education Sharing Platform in Higher Vocational Education   Order a copy of this article
    by Yuchen Song, Yi Chen 
    Abstract: The large amount of shared resources has brought difficulties to learners’ choices, making personalised learning difficult. Therefore, the characteristics of learning data related to electronic education platforms were analysed in the experiment. A student performance monitoring model was constructed through improved deep confidence networks (DBNs), linear regression, and other methods, and student performance was evaluated. Through experiments, in the loss performance test of student performance monitoring model, the improved DBN model has better convergence effect and loss performance than BP and DBN models. In the prediction of student performance, the prediction accuracy of the improved DBN model is 0.956, while that of BP model and DBN model is 0.765 and 0.864, respectively. The comprehensive performance of the improved DBN model is the best. The research content has important reference value for Digital transformation of higher vocational education.
    Keywords: learning characteristics; electronic education sharing platform; ESP; depth confidence; individual learning; test.
    DOI: 10.1504/IJCSYSE.2025.10059047
  • The Algorithm of College Students' Physical Fitness Analysis Based on Data Mining Under the Background of Online Courses   Order a copy of this article
    by Yi Lu  
    Abstract: To promote the healthy advancement of college students' physical fitness and accurately and scientifically analyse the students' physical fitness test results, this research proposes a data mining based college students' physical fitness analysis algorithm. The physical fitness data of colleges, classes, and individuals is analysed, and the corresponding suggestions are proposed. The results show that there are 5128 valid data points for boys and 7812 valid data points for girls obtained by optimising K-means. In the case study, radar analysis was used. The results showed that the students' standing long jump was the strength of the class, and the lower limb strength developed well; The rest are all yellow areas, which shows that their development is balanced. If the score of the students' 50-metre run is in the orange area, some special training can be carried out.
    Keywords: data mining; decision tree; physical fitness; optimisation algorithm; platform design.
    DOI: 10.1504/IJCSYSE.2025.10059069
  • Analysis of the digital technology of the e-education platform in practice in tertiary education   Order a copy of this article
    by Fei Gu  
    Abstract: The study collects learners' learning preferences with the help of deep belief networks, allocates and updates course resources with BP neural networks, and introduces the concept of similarity to effectively connect learner resources with teaching resources to better meet learners' cognitive level and logical thinking. The performance of the proposed algorithm model is tested and the results show that the DBN-BP algorithm achieves minimum RMSE and MAPE values of 0.66 and 0.63 in terms of recommendation performance, and achieves 95.23% and 0.72 in terms of resource recommendation accuracy and coverage, effectively improving the teaching recommendation performance. The algorithm can effectively provide new and improved ideas and means for online practical teaching in higher education, and provide guarantee for its teaching quality improvement.
    Keywords: e-education platforms; digital technology; higher education; deep belief networks; backpropagation; BP; similarity.
    DOI: 10.1504/IJCSYSE.2025.10059070
  • Building a collaborative university teaching platform based on online social space   Order a copy of this article
    by Xiaomei Zhang  
    Abstract: To recommend learning resources that are suitable for students' learning development, a content recommendation algorithm and KRCF algorithm are proposed based on the relationship between students and knowledge resources, as well as knowledge similarity user groups, to correct and predict students' learning paths, and to establish a university teaching collaboration platform. The experimental results show that the KRCF algorithm performs better than algorithms such as the CF algorithm. Under different iterations or training set ratios, its RSME and MAE values are the smallest. When the training set ratio is 0.15, its MAE value is 0.72, which is lower than the MCS-CF algorithm. Overall, the accuracy of the KRCF algorithm is over 70%, with an accuracy of 94.67% when there are 50 neighbours. The accuracy of content recommendation algorithms in recommending learning resources is high, and personalised systems can quickly respond to user requests.
    Keywords: online social space; content recommendation algorithm; knowledge association; collaborative filtering recommendation algorithm.
    DOI: 10.1504/IJCSYSE.2024.10058154
  • Research on privacy protection method based on deep reinforcement learning algorithm in data mining   Order a copy of this article
    by Yan Cai, Rui Xue 
    Abstract: Protecting data privacy is a critical issue in information security. However, traditional methods often hinder data mining efficiency and accuracy. This study aims to balance data security and mining efficiency to improve accuracy while ensuring privacy. Using the deep reinforcement learning algorithm, the target-optimised deep Q-network (T-DQN) is proposed. Multiple standard network datasets are used for testing. The proposed algorithm achieves higher Bayesian network representation accuracy (7.2%24.7%) compared to PrivBayes under different datasets. Revenue is also increased (12%17% higher than Sarsa and 29%32% higher than Q-learning). Weak regret value is lower (6982,573 lower than Sarsa and 9841,327 lower than Q-learning). The algorithm demonstrates good convergence, adaptability, and superior performance compared to other algorithms. It provides a reference for improving privacy protection efficiency in data mining.
    Keywords: data mining; deep learning; privacy protection; DQN; Bayesian network; target-optimised deep Q-network; T-DQN.
    DOI: 10.1504/IJCSYSE.2024.10058272
  • TQ Evaluation of Higher Vocational English Sustainable Education Based on GA Algorithm and BP Algorithm   Order a copy of this article
    by Yanlei Ma, Yuxia Zheng 
    Abstract: In higher vocational (HV) colleges English teaching (ET), the improvement of teaching quality (TQ) evaluation means has a certain promoting effect on improving TQ. To effectively evaluate the quality of ET in HV colleges, this paper uses GA algorithm to improve convergence effect and shorten training time on the basis of BP algorithm. In the application analysis, the evaluation results of the method used in the article are less different from the students' evaluation results. The evaluation difference in the sample set is less than 4.000%, and the minimum test evaluation error is 0.164%. The minimum test evaluation error of BP algorithm is 5.126%. And the accuracy rate of TQ evaluation of the method used in the article reaches 98.84%, which is 6.23% higher than that of BP algorithm. The method adopted in the article can be applied in vocational ET, which is conducive to improving the efficiency and accuracy of TQ evaluation. And it has a positive effect on improving the quality of vocational English education.
    Keywords: BP algorithm; sustainable education; TQ; HV English; GA algorithm.
    DOI: 10.1504/IJCSYSE.2025.10059071
  • Improving Emotion Classification of the Hotel’s Customer Satisfaction Online Reviews: A Self-Adaptive Ensemble Approach   Order a copy of this article
    by Xin Li, Ding-Bang Luh, Zi-Hao Chen, Yue Sun, Guanyu Pan, Qianer Li 
    Abstract: With the evolution of the tourism industry, updating the emotion classification model using artificial intelligence approaches become essential. This article aims to address the issue that hotel online reviews of customer satisfaction are affected by many attributes. To test the proposed framework, this article utilised over 1,483 online reviews of Hainan Island hotels on to extract accommodation factors and sentiment terms. The TF-IDF term weighting integration module was then applied. Finally, a self-adaptive ensemble model for criticism was built based on the existing database. This study provides an example of how machine learning models can be applied to improve hotel accommodation service quality. Moreover, the differentiation of positive and negative comments by artificial intelligence tools allows for the handling of countless statistics that were previously impossible. The research approaches are significant for studying customer satisfaction, particularly in the context of the tourism industry's economic recovery.
    Keywords: text mining; self-adaptive ensemble model; hotel service development; online reviews; customer satisfaction.
    DOI: 10.1504/IJCSYSE.2024.10058290
  • Research on the BP-Network Education Assessment System for Industrial Integration Education Strategy in University Economics and Management Majors   Order a copy of this article
    by Yifei Wang  
    Abstract: To address the insufficient educational assessment methods in industrial integration education strategy for economics and management majors, a BP-network assessment pattern was designed. This pattern utilised the analysis of relevant literature and student questionnaires to establish an assessment system and an index system. Combining the index system with the model architecture, an assessment model for industrial integration education strategy in economics and management majors was established. The model presented a maximum error of 0.97, indicating stability, and it showcased a better fitting effect in overall fitting. Based on this, the model evaluation results show the education focus directions with scores above 0.9 and below 0.6, which can help universities maintain the strengths of the existing education programs and suggest the weaknesses of the education programs, and promote universities to improve the weaknesses. The results of this study can provide a theoretical basis for promoting educational reform in niche areas.
    Keywords: economic management; Industrial integration education strategy; educational assessment; BP-network; weight matrix.
    DOI: 10.1504/IJCSYSE.2025.10060230
  • Research on the Application of English Online Course Recommendation Model Based on Machine Learning in College English Teaching   Order a copy of this article
    by Qiongying Sun  
    Abstract: The lack of text information, user behaviour information, and evaluation information in English online courses result in traditional recommendation algorithms not being directly applied to course recommendations. To solve this problem, based on the full analysis of NetEase cloud class user data, the study extracted four types of characteristics, namely user preference characteristics based on topics and collaborative filtering (CF), course popularity, and course instructor influence, and quantified these characteristics that affect recommendation decisions. Then, the ranking SVM algorithm was used to sort the multi class features obtained, and a multi feature network course recommendation model was constructed, which summarised the recommendation problem as a sorting problem. Finally, user interest labels are obtained using topic based preference features. Through the above operations, a machine learning (ML) based online English course recommendation model was constructed. Through experimental analysis, it can be seen that the average precision of Model 1 is 87.86%, the average recall value is 79.57%, the average RMSE value is 0.265, and the average MAE value is 0.285. The research and construction of the model can provide students with more accurate and personalised intelligent course recommendation lists when using online classrooms for learning.
    Keywords: online education; machine learning; ML; course recommendation; characteristic sorting.
    DOI: 10.1504/IJCSYSE.2025.10060231
  • Wwireless mobile technology in innovative teaching in universities   Order a copy of this article
    by Min Li  
    Abstract: As wireless mobile technology develops quickly; education has ushered in new development. Based on wireless mobile technology, smart classrooms are becoming more prevalent in modern schooling. Personalised learning, which breaks through the limitations of time, place, methods, and learning resources. At this stage, with the support of wireless mobile technology, allowing students to conduct personalised learning according to their own interests is the focus of reforming education and teaching. The existing personalised learning resources for students have certain defects in the mining of students' interests, so there are deficiencies in the recommendation of learning resources. Aiming at this problem, through the K-means clustering, a recommendation model for students' personalised learning resources is established. An experiment is conducted using the online learning course data of a university student. Experimental findings indicate that the recommendation model has an accuracy rate of 99.23%. This outperforms the K-means model and the KCF model by 6.28% and 4.06% respectively. Therefore, the improved collaborative filtering recommendation algorithm based on K-means proposed in the study has a good recommendation effect. It can effectively explore students' interests and recommend corresponding learning resources to meet their personalised development needs.
    Keywords: wireless mobile technology; innovative teaching; learning resource recommendation; K-means; collaborative filtering.
    DOI: 10.1504/IJCSYSE.2025.10060232
  • A study on the optimization of university English teaching based on an enhanced decision tree model in the context of big data   Order a copy of this article
    by Haiyan Cai 
    Abstract: The purpose of learning is the most important factor affecting an individual’s percentage of making notable progress (PMNP). In the context of big data, this study selected a sample of 1,805 non-English majors in a university to investigate the optimisation of university English teaching based on an enhanced decision tree model. The results showed that among the 1,805 student sample, 352 students made more significant improvements in their test scores compared to the last test, accounting for 19.50% of the total PMNP of the entire sample. The Chi-squared automatic interaction detector (CHAID) decision tree model was used to identify implicit and valuable factors influencing teaching quality based on data on the process, conditions and environment of English language teaching. The results show that through calculation of CHAID decision tree, the resultant data of each node is a reflection of the effect of each factor on PMNP.
    Keywords: big data; CHAID decision tree; English language teaching; percentage of making notable progress; PMNP; genetic algorithm.
    DOI: 10.1504/IJCSYSE.2025.10061164
  • Design of a platform for evaluating the effectiveness of college and university English education by integrating PSO and SVM   Order a copy of this article
    by Dan Xu  
    Abstract: The study established that PSO is used to enhance the optimal solution of the disk operation parameters and SVM regularisation parameters. The results demonstrated that the effect evaluation model constructed by PSO-SVM obtained the best adaptation 100 at five iterations and reached the fastest convergence speed. Among different model errors, the data error distribution of the PSO-SVM pattern is closest to 0, with the smallest absolute error variation and the highest prediction accuracy; in practical applications, the classification precision of PSO-SVM is as high as 97.01%, which is much This is much higher than the outcomes of the BP and SVM models. The outcomes implied that the evaluation platform for English education teaching effectiveness evaluation under this model had the best evaluation effect, can achieve low evaluation time and high evaluation accuracy both at the exact moment, and offers a brand-new reference for teaching evaluation.
    Keywords: PSO; SVM; English language teaching; evaluation platform.
    DOI: 10.1504/IJCSYSE.2025.10060233
  • Research on Key Technologies of Privacy Protection in Big Data Computing Environment   Order a copy of this article
    by Jiaming Nong, Mengzhen Chen 
    Abstract: In terms of big data computing, it is mainly based on data sharing, which includes various information technologies. Although it provides corresponding assistance for people to solve problems, there are also privacy breaches. In order to improve the security of privacy, people have carried out corresponding encryption, authentication, authorisation and other operations on privacy, but the problem of information leakage has not been effectively solved, posing a threat to people's privacy security. In big data computing, there are relatively many factors that lead to privacy breaches. In order to ensure people's privacy and security, this article conducts research on key technologies of privacy protection based on previous research by collecting a large amount of literature.
    Keywords: big data; data security; privacy protection; key technology.
    DOI: 10.1504/IJCSYSE.2024.10058676
  • The application of big data platform in intelligent scheduling of tourism passenger flow based on time window   Order a copy of this article
    by Yanhua Guo 
    Abstract: To manage rural leisure tourism tourists reasonably and achieve intelligent guidance, this paper adopts intelligent scheduling of passenger flow, combining Wi-Fi detection technology and time-window-based analysis method for passenger flow data collection and analysis. Based on the booster and gravitational functions, the corresponding gravitational scheduling model is proposed with comprehensive consideration of multiple factors to perform passenger flow scheduling, and the related supervisory scheduling system is designed. The results show that after analysing the selected real-time passenger flow and time-load rate of the leisure village, the MFGM method adopted in this paper can effectively dispatch the passenger flow and make the time-load rate of the leisure village converge to 1. The average satisfaction and the average time-load rate after diversion of the MFGM method are 0.640 and 0.839, respectively. The dispatching scheme used in this paper can effectively reduce the peak passenger flow of the leisure village and possesses better practical value.
    Keywords: time window; scheduling; visitor diversion; Wi-Fi; gravitational scheduling model.
    DOI: 10.1504/IJCSYSE.2024.10058810
  • Design of music signal enhancement system based on big data clustering technology   Order a copy of this article
    by Taoru Kong, Yanli Shen 
    Abstract: Electronic music signals are prone to distortion due to noise interference. Therefore, an intelligent music signal enhancement system based on big data technology is designed. The system hardware consists of music signal acquisition module, audio processing module and audio codec module. It can extract the features of music interference signals based on big data clustering algorithm and use improved wavelet transform to denoise music signals. The music signal intensifier in the model adopts the autocorrelation filtering algorithm to filter the separated signals. This algorithm can separate the noiseless music signal from the noise signal. Finally, the test set music signals are input into the neural network. After the multi-dimensional feature vector of the signal passes through the hidden layer and the output layer, the class number of each music signal can be obtained and the automatic recognition of music signals can be realised. Experimental results show that the proposed method can reach 30.12 dB, the average gain coefficient is 0.987, and the bit error rate is 6.81%. The noise reduction performance and music signal enhancement effect are superior to the other two methods. Therefore, this method has certain practical value and application prospect.
    Keywords: big data clustering; audio processing; improved wavelet transform; music signal enhancement.
    DOI: 10.1504/IJCSYSE.2024.10058811
  • Construction of an intelligent recommendation model for microblog short videos based on ABC-BPNN   Order a copy of this article
    by Yunpeng Gu  
    Abstract: Crowdsourcing testing is an emerging method that improves software product quality by quickly identifying and rectifying defects. This study optimises the application of clustering algorithms in crowdsourcing testing to enhance test report review efficiency. It demonstrates that the best defect detection is achieved at a 0.4 threshold, and a 0.7 weight yields the optimal harmonic average. The clustering effect is best when text information is primary, and screenshot information is secondary. The proposed model improves upon traditional models by 9.336%. The high-precision information processing algorithm proposed herein considers the interrelationships in test reports, thereby improving the accuracy of the clustering algorithm. This enhances the work efficiency of auditors, testing quality, and reduces testing costs.
    Keywords: back propagation neural network; BPNN; bee colony intelligence algorithm; short video; principal component analysis; intelligent recommendation.
    DOI: 10.1504/IJCSYSE.2024.10058812
  • Application of New Media Technology in Opera Networked Teaching   Order a copy of this article
    by Yan Chen  
    Abstract: Modern new media technology provides a good platform for education and teaching. The use of new media technology to assist online teaching of traditional Chinese opera has shown great potential for development in major universities. This study designed a networked teaching mode for traditional Chinese opera and new media technology and improved LSTM using CTC to construct a BiLSTM-CTC speech recognition model. The results show that for 8 and 12 rounds of training, the speech recognition error rates of the proposed BiLSTM-CTC model are 60% and 7%, respectively. In college opera teaching, the minimum value of this model is only 6%, the consumption time is 17.38 seconds, and the comprehensive score is as high as 95.97 points. The above results indicate that this method can accurately identify voice features in traditional Chinese opera teaching, providing technical and methodological references for effective interaction in the online teaching of traditional Chinese opera.
    Keywords: new media; opera; networked teaching; long short-term memory; LSTM; speech recognition.
    DOI: 10.1504/IJCSYSE.2024.10058859
  • Wavelet analysis-based techniques for processing voice and video signals in network communications   Order a copy of this article
    by Fangshu Liu  
    Abstract: Aiming at the shortcomings of traditional wavelet analysis techniques in signal processing, the study proposes an improved threshold wavelet analysis method to optimise its signal denoising effect. Firstly, the important components of wavelet analysis theory are discussed. Secondly, based on the soft threshold and hard threshold functions for signal denoising, an adjustment factor is introduced to optimise the convergence for communication signals smaller than the threshold. The outcomes demonstrate that the improved wavelet thresholding approach efficiently lowers noise while preserving the precise properties of the original signal. The technique generates a reconstructed signal with lesser distortion and fewer noise residuals when compared to previous approaches. The signal-to-noise ratio of the improved thresholded wavelet output is consistently over 90, and the distortion deviation is very near to 2. In conclusion, this method of communication signal processing can enhance communication quality while reducing interference from the outside world.
    Keywords: network communication; signal processing; wavelet analysis; modulation factor; threshold function.
    DOI: 10.1504/IJCSYSE.2024.10059073
  • Research on the Application of Clustering Algorithm Based on Hybrid Similarity in High Precision Information Processing of Crowdsourcing Test Reports   Order a copy of this article
    by Li Huang, Xin Zheng 
    Abstract: Crowdsourcing testing is an emerging method that improves software product quality by quickly identifying and rectifying defects. This study optimises the application of clustering algorithms in crowdsourcing testing to enhance test report review efficiency. It demonstrates that the best defect detection is achieved at a 0.4 threshold, and a 0.7 weight yields the optimal harmonic average. The clustering effect is best when text information is primary, and screenshot information is secondary. The proposed model improves upon traditional models by 9.336%. The high-precision information processing algorithm proposed herein considers the interrelationships in test reports, thereby improving the accuracy of the clustering algorithm. This enhances the work efficiency of auditors, testing quality, and reduces testing costs.
    Keywords: crowdsourcing testing; image segmentation; text segmentation; hybrid similarity; clustering algorithm.
    DOI: 10.1504/IJCSYSE.2024.10058863
  • Big data analysis for sustainable music teaching using collaborative filtering recommendation optimization algorithm   Order a copy of this article
    by Lanqian Liu  
    Abstract: The use of big data analysis technology to meet the needs of sustainable development of music teaching has become a current research hotspot. In order to improve the teaching effect of students and teachers, this study first proposed the K-means clustering algorithm (BSLCSK means) based on two-way selective learning chicken flock optimisation (BSLCS), and then designed the improved cooperative filtering algorithm (BSLCSIUCF) integrating BSLCSK means. The results show that the maximum and minimum mean absolute errors (MAE) of the BSLCSIUCF algorithm are 1.1 and 0.4, respectively. When the number of neighbours is 10 and 20, the MAE values of the algorithm are 1.1137 and 0.9144, respectively. The accuracy of prediction and recommendation is further improved. While obtaining more accurate recommendation results, the recommendation results obtained are also more stable, providing better music big data teaching services for college teachers and students.
    Keywords: K-means clustering; chicken swarm algorithm; collaborative filtering algorithm; music; big data.
    DOI: 10.1504/IJCSYSE.2024.10058864
  • Application of information dissemination network embedding in social media user classification and influence analysis   Order a copy of this article
    by Jia Li, Jiangnan Li 
    Abstract: To analyse the spread of influence among users, a method based on information network embedding is proposed. During the process, not only was a user classification method based on heterogeneous information network embedding proposed but also a model of user influence intensity and influence propagation was constructed using information dissemination network embedding technology. The results showed that in the comparison of classification performance among different methods, the accuracy, recall, and macro F1 values of the research methods reached 0.823, 0.867, and 0.885, respectively; When the ratio of seed users to affected users is 1.6%, the research method has the highest number of affected users, up to 13.8%. At this time, the research method has a shorter time for classifying users. The above results indicate that the research method has relative superiority, high accuracy and feasibility in user classification, and can achieve low runtime and high classification accuracy.
    Keywords: information dissemination; web embedding; user classification; influence analysis.
    DOI: 10.1504/IJCSYSE.2024.10058865
  • Research on network intrusion detection model that integrates WGAN-GP algorithm and Stacking learning module   Order a copy of this article
    by Xiaoli Zhou  
    Abstract: With the development of network technology, current network intrusion detection models have effectively detected some network intrusion methods. In order to improve the detection performance of network intrusion detection models, a new network intrusion detection model combining data augmentation technology is proposed. The model incorporates the WGAN-GP data augmentation module for data balance enhancement and a stacking learning module for model classification accuracy. In the performance comparison analysis of the WGAN-GP algorithm, it was found that the accuracy and F1 value of the WGAN-GP algorithm were 98.25% and 0.792, respectively, which were superior to the comparison algorithm. The above results indicate that the detection performance of the WGAN-GP algorithm is superior to that of the comparison algorithm. Therefore, integrating the WGAN-GP algorithm into network intrusion detection models can effectively improve its intrusion detection performance and promote the development of the field of network intrusion detection.
    Keywords: stacking algorithm; WGAN-GP algorithm; network ID model; WGAN algorithm; SMOTE algorithm; ADASYN algorithm.
    DOI: 10.1504/IJCSYSE.2025.10059074
  • Application of CGE model-based digital media technology in virtual reality project design   Order a copy of this article
    by Kaiming Wang  
    Abstract: With the rapid development of the internet, the size and scale of the digital economy are constantly undergoing reform and upgrading. The virtual reality project the digital economy, as a newly emerging economic form in the current era, needs to clarify the theory of digital economy industry development under the digital economy, and develop the digital economy in combination with local characteristics. This study proposes the application research of digital media technology in the digital economy based on the CGE model. Firstly, a CGE model is constructed, and time attributes are added to the CGE model to form a dynamic CGE model. And analyse this, this method helps to calculate the contribution of the digital economy industry. Afterwards, the model was simulated and validated. From 2019 to 2022, the maximum absolute error of the dynamic CGE model was around 3.89% in terms of data accuracy. The simulated values of this model have a small difference from the actual values in national statistics, with absolute error values less than 4.98%. Meets statistical accuracy requirements for data. Experimental analysis shows that the construction of the model is basically in line with reality.
    Keywords: CGE model; digital economy; virtual reality project; digital media technology.
    DOI: 10.1504/IJCSYSE.2024.10058994
  • Machine learning based user profile recognition for popular short videos on social platforms   Order a copy of this article
    by Ying Shi  
    Abstract: Considering the emotion and opinion tendentiousness contained in user comment data, a user comment data mining model is based on bidirectional encoder representation from transformers bidirectional long short term memory conditional random field (BERT-Bi LSTM-CRF). BERT-Bi LSTM-CRF is designed to obtain text sequence features and annotation terms. In addition, to recognise the emotional polarity of aspect words, a classification model based on BiAtt-GCN is constructed. For the dataset SemEval Task 4, the proposed model achieved an accuracy improvement of 0.83% and 3.2%, respectively, compared to the BiSLTM CNN model and CMLA model, and an increase of 0.68% and 1.26% in the recall index. Therefore, the model proposed in the study is effective in the analysis of user comment data.
    Keywords: short video; aspect words; BERT-Bi LSTM-CRF; user portrait; classification.
    DOI: 10.1504/IJCSYSE.2024.10058996
  • Vocational College Employment Training and Career Planning Model Design Based on Improved Collaborative filtering   Order a copy of this article
    by Jin Wang  
    Abstract: In the manuscript, a vocational college student employment training and career planning model based on collaborative filtering is proposed to recommend suitable employment training and career planning for students. Focusing on the flaws of collaborative filtering algorithm in data mining of students’ employment behaviour, the K-means clustering algorithm and Kruskal are combined to optimise it. The experiment is conducted using the employment data of vocational school graduates in the past three years. The outcomes indicate that the accuracy of this model reaches 94.18%, which is 2.93% and 2.12% higher than that of CF and KCF respectively. It proves that this method can basically meet the career planning and vocational training needs of vocational school students in the employment process, establishing a good connection between students and enterprises.
    Keywords: vocational colleges; employment training; career planning; collaborative filtering; Kruskal algorithm.
    DOI: 10.1504/IJCSYSE.2025.10060234
    by Jiafei Geng, Xiaoli Wu 
    Abstract: The research firstly improves the keyword extraction algorithm in the text mining algorithm by introducing information entropy, relative entropy, word length weighting factor and word location weighting factor, and processes them in the Spark computing framework. The improved algorithm was then applied to a web text mining system and used to mine and collect information from internet customers. To test the performance of the proposed algorithm, the study compared the recall, accuracy and F-values of the four algorithms under four datasets. The proposed algorithm was found to have better performance with maximum recall value, accuracy value and F-value of 81.6%, 84.8% and 84.1% respectively in the dataset. Finally, it was found that the latter could achieve a maximum prediction accuracy of 99.5%, which is much more accurate than the traditional algorithm for customer information collection.
    Keywords: text mining; web; information gathering; big data; computational framework.
    DOI: 10.1504/IJCSYSE.2024.10059171
  • Research on the continuous improvement mechanism of computer major practice teaching in applied universities   Order a copy of this article
    by Lin Zhu, Li Zhuang 
    Abstract: A set of continuous improvement mechanism is constructed for the problems of simplification of practical teaching objectives, detachment of teaching contents from reality, single form of teaching, and popularisation of teaching methods in computer science. Firstly, this paper takes result-oriented education as the theoretical basis. Secondly, the literature research method and case study method are used to take evaluation to feedback to continuous improvement as the entry point. Further, the continuous improvement mechanism is built around the computer science major from basic and skill experiments, comprehensive and improvement experiments, design and innovation experiments, and technology and research experiments. Finally, the evaluation mechanism, feedback mechanism and continuous improvement mechanism of computer practice teaching are explained in detail with actual cases to make the proposed mechanism more practical.
    Keywords: computer science; outcomes-based education; OBE; major; practical teaching; continuous improvement.
    DOI: 10.1504/IJCSYSE.2024.10059202
  • A Security protocol with trust model for WSN   Order a copy of this article
    by Ruizhi Chen  
    Abstract: The research introduces a novel secure routing protocol for wireless sensor networks by integrating a trust model. The protocol incorporates acceleration factors and penalty coefficients to enhance the trust mechanism, enabling faster identification of malicious nodes and swift reduction of their trust values. The protocol introduces the acceleration factor and penalty factor into the trust mechanism to increase its response speed and trust value drop rate. Performance tests reveal that the protocol exhibits over 60% faster trust value reduction speed than TLEACH and 37% faster than the TLES protocol under black hole attacks. Additionally, the protocol demonstrates a longer life cycle, with the first node expiring around round 422 and half of the nodes ceasing operation by round 537, surpassing the longevity of TLEACH and TLES. These findings highlight the protocol’s promising practical application potential.
    Keywords: wireless sensor network; WSN; security protocols; network structure; network security.
    DOI: 10.1504/IJCSYSE.2025.10059616
  • Research on Short Text Information Mining and Classification Methods for Social Media   Order a copy of this article
    by Tingting Wang  
    Abstract: The internet era has brought massive information output and dissemination, and social media, mainly represented by WeChat and Weibo, have gradually become an important part of people's daily life. As the data of short texts generated by social media are growing, how to extract and classify useful information from these texts has become a pressing problem. The study designs a co-occurrence information model to build a graph structure of short texts and classifies them by combining a graph convolutional network and introducing an attention mechanism. The outcomes demonstrate that the precision of the upgraded model is 82.94% and 90.03% in the datasets MR and HR, respectively, with better classification outcomes; the precision of the model is basically stable at 80% and above, up to 90% under the change of training data size; the error rate is only 8.66% and the time required is 29.85% in the classification of short textbooks in microblogging platform. The precision and operational efficiency provide a new technical and methodological reference for the information processing of social media.
    Keywords: social media; short text; information classification; graph convolutional network; attention mechanism.
    DOI: 10.1504/IJCSYSE.2024.10059265
  • Analysis of Online Agent Accounting Platform under Big Data AI Environment   Order a copy of this article
    by Zhifeng Yuan 
    Abstract: This paper aims to provide a comprehensive overview of the current status and future prospects of the agent bookkeeping service industry in China, focusing on the online agent bookkeeping platform developed by Nanjing Cloud Accounting Network Technology Co., Ltd. By analysing the business process of this platform, the paper aims to highlight the benefits of an online approach to agent bookkeeping, such as improved work efficiency, accuracy, reliability, scalability and flexibility. The comparison with traditional agent bookkeeping also demonstrates the potential advantages of adopting innovative approaches in the industry. The insights gained from this analysis provide valuable direction for the future prospects of the agent bookkeeping service industry in China.
    Keywords: online service platform; agent bookkeeping; big data; AI.
    DOI: 10.1504/IJCSYSE.2024.10059275
  • Research on End-to-End Computing Power Network Architecture Based on SDN and MIH Technology   Order a copy of this article
    by Fang Cui, Jianhua Gu, Hengjiang Wang, Mao Ni, Ting Zhou 
    Abstract: For ubiquitous access, the study selects software-defined networks with independent switching, control upgrades, and centralised management using the manager to achieve the same control. Pre-filtering and D-VIKOR are used in network switching decisions to help perform network decisions. The experimental results show that the network architecture is suitable for mobile network scenarios. In the 30-m/s in-vehicle scenario, the latency is reduced by 27.55%, and the overall switching effectiveness is improved by 10.21%. For the situation that the increase in arithmetic power will lead to more nodes, which will increase switching errors and blocking, the relevant test results show that the number of vertical switching errors for the SDN architecture is less than that of the D-TOPSIS architecture 285 times less. This indicates that the end-side network architecture constructed in the study can handle the switching requirements of heterogeneous networks and also shows excellent performance on different problems in multiple scenarios.
    Keywords: count-network convergence: end-side network; architecture; switching delay; mobile scenario.
    DOI: 10.1504/IJCSYSE.2025.10059362
  • Interactive Piano Automatic Accompaniment Intelligent System Based on Machine Learning Model   Order a copy of this article
    by Wei You  
    Abstract: In order to enrich the expression of piano melodies, the article applies the IBi-LSTM algorithm for the automatic arrangement of piano harmonies. By constructing a network scoring platform for users to audition and score. The results show that the IBi-LSTM algorithm performs better and has less perplexity than algorithms such as LSTM. Compared with other methods, the multi-basic frequency estimation method used is more effective, with a higher recall of 84.42% and a higher F-value of 81.38% under the MUS subset. In the harmonic arrangement effect, most of the auditioners rated higher than 4, with an average rating of 3.99 and a maximum rating of 4.6. The article uses the method to achieve automatic piano accompaniment and is well received by the auditioners.
    Keywords: Bi-LSTM; machine learning; interactive; piano; accompaniment; harmonic arrangement.
    DOI: 10.1504/IJCSYSE.2025.10060237
  • Research on GCN-based Aerobics Movement Recognition under the Background of Big Data   Order a copy of this article
    by Li Shang  
    Abstract: The effective recognition of aerobics is a powerful guarantee to predict athletes’ physical injury in time, and also a powerful tool to improve the standard level of aerobics. To solve the problem of low nonlinear ability of spatiotemporal graph Convolutional network (STGCN) model, a dense connected network structure based on STGCN is proposed in this paper. Finally, an aerobics recognition model is constructed by combining the dense spatiotemporal graph Convolutional network (DSTGCN) algorithm. The integrated model first analyses and preprocesses the images in aerobic exercise videos to construct a directed spatiotemporal map of human bones. The resulting skeleton topology is then embedded in the DSTGCN network, where it is learned and updated along with the model. The experimental results show that the final recognition accuracy of DSTGCN model is stable at about 86.95%, which is better than other existing algorithms.
    Keywords: big data; STGCN; DSTCGN; directed space-time graph; aerobics exercise; motion recognition.
    DOI: 10.1504/IJCSYSE.2025.10059494
  • Construction of a Computer-Assisted English Language Testing Model   Order a copy of this article
    by Honglei Zhu 
    Abstract: Many large English tests in China have gradually completed the transformation from the traditional mode to computer intelligence, but many problems still exist. This paper takes the English major students of a school as the research object, and discusses the influence of traditional mode and computer aided method on students’ language proficiency test results. By studying the comparative data of English majors in a teacher training university, this paper concludes that: the overall speed of marking is optimised at more than 81%, which greatly saves the time spent on marking, reduces operational errors, greatly improves the efficiency of marking, and reduces the difficulty of marking work; it can be seen that the application of computer-aided technology brings a great improvement to the overall reliability and validity of English speaking tests At the same time, it can be seen that the application of computer-aided technology has greatly improved the overall reliability and validity of the oral English test, making the test more objective, fair and rigorous, and more credible; the satisfaction rate of students is 81.5% and that of teachers is 60%.
    Keywords: computer-aided technology; English language testing; data mining; DM; online analytical processing; OLAP.
    DOI: 10.1504/IJCSYSE.2025.10059495
  • Research on the Importance of Database System Security Performance Testing Technology to Computer Software Development   Order a copy of this article
    by Li Gao, Qiu Junlin, Huaqi Lu, Xiaolin Jiang, Shaohang Yi 
    Abstract: In order to make up for the lack of database security in traditional business systems and help managers realise real-time monitoring and auditing of database operations, the research is based on cloud-encrypted databases and combined with auditing technology to ensure database security and integrity during software development sex. In the study, two normal users and two abnormal users are used to operate the database to judge the effectiveness of the security audit scheme. The results show that the judgment value can distinguish normal users from abnormal users, and false detection rate can be kept below 0.03. As length of sliding window increases, detection rate of system shows an increasing trend, and the false detection rate shows a decreasing trend, but the change gradually slows down, and the impact on it decreases significantly after reaching 0.015. It should be noted that with the increase of the sliding window, the time complexity of the system is also increasing, which will have a certain impact on real-time performance of audit. The research results ensure that database managers can discover the existing problems in the first time, and formulate targeted solutions to improve the efficiency of software development.
    Keywords: big data; cloud computing; hidden Markov model; HMM; encrypted database; security audit.
    DOI: 10.1504/IJCSYSE.2025.10059710
  • Two-way Sentiment Analysis Method of Multimedia Information Based on Deep Learning Algorithm   Order a copy of this article
    by Yingjie Liu, Baopeng Kan 
    Abstract: Emotional analysis can better understand the public's emotions and needs, and make better decisions based on the analysis results. However, there is still a lack of effective analysis methods in practical applications. Therefore, the study utilises a bidirectional emotion classification mechanism based on deep learning, and uses time series algorithms to predict the development trend of users' bidirectional emotions. A bidirectional emotion analysis method for multimedia information based on deep learning algorithms is proposed. The results show that the accuracy of emotional judgment in the analysis model is 82.5%, which is 11.1% higher than the machine learning model. At the same time, the prediction accuracy of the prediction model is around 84%, which is significantly better than the comparison method. This indicates that the bidirectional emotion model constructed through research can accurately analyse user emotions and provide reference for making development decisions in the multimedia field.
    Keywords: deep learning; two-way sentiment analysis; attention mechanism; time series; sentiment prediction.
    DOI: 10.1504/IJCSYSE.2025.10059764
  • The Influence of Knowledge Management Mode of Vocational Education based on Information Technology on Students' Learning Ability   Order a copy of this article
    by Tian Xie  
    Abstract: At present, the application of informatisation in vocational education is not mature, and its impact on teaching activities is not clear. Therefore, this study has carried out research on the KM model built by information technology. To explore the impact of the knowledge management (KM) model constructed this time on students’ learning ability (SLA), a new evaluation index system was constructed. By RSBP neural network, the evaluation model was constructed. In the result analysis, the research tested the performance of the model, and verified the impact of information KM on SLA through relevant analysis. Rough set can effectively improve the accuracy of BP neural network evaluation, and the evaluation error of RSBP was only 0.07 in the simulation experiment. The correlation showed that information KM had a significant positive correlation with SLA. This study provided guidance for the application of information KM in vocational education and had good reference value.
    Keywords: informatisation; vocational education; knowledge management; KM; evaluate; students’ learning ability; SLA.
    DOI: 10.1504/IJCSYSE.2025.10059919
  • Application Value of Data Mining Technology in Ultra Dense Heterogeneous Wireless Networks   Order a copy of this article
    by Yuming Zhong, Leyou Chen 
    Abstract: In the era of the internet, a large amount of data is constantly generated, which has led to the emergence of network data mining technology. To improve access network security and user network experience, data mining technology is applied to ultra dense heterogeneous wireless networks. A switching algorithm based on user personalised preferences is proposed and a network security prediction module based on data mining is designed. Experimental data shows that when the number of networks is 10,000, the computational time cost based on the multi-attribute vertical switching algorithm is 3.45 ms. The switching algorithm based on user consumption preferences has a computational time cost of 0.97 ms, saving approximately 71.9% of the time. When the number of users exceeds 200, the throughput of the predictive network security switching algorithm based on data mining exceeds that of the analytic hierarchy process switching algorithm. The blocking rate is lower, which can better achieve balanced network selection and improve user network experience.
    Keywords: data mining; ultra dense isomerism; wireless network; user preferences; network security.
    DOI: 10.1504/IJCSYSE.2025.10059939
  • Feature Perception Based Graphic Advertising Image Generation Technology   Order a copy of this article
    by Huichao Zhang  
    Abstract: In order to meet the market demand for graphic advertising images, this article proposes a feature aware image generation technology for print advertising. This technology quantifies image features, uses simulated annealing algorithm to sample the quantised features, and then combines dictionary strategy to optimise probability models to predict feature distribution, ultimately generating the optimal graphic advertising image. The results show that in terms of iteration error rate, the simulated annealing algorithm tends to stabilise after 85 iterations, with an error rate of 0.015. In terms of colour feature extraction rates, the simulated annealing algorithm has extraction rates of 92%, 91.5%, and 89.1%, respectively. In expert evaluation, the expert evaluation scores all exceed 90 points. The above data indicates that the proposed method is feasible and can provide technical support for related advertising image generation.
    Keywords: graphic advertising; image generation; simulated annealing algorithm; lexicographic strategy; feature perception.
    DOI: 10.1504/IJCSYSE.2025.10060143
  • Application of Electronic Information Technology Based on Optical Sensors in Intelligent Transportation Systems   Order a copy of this article
    by Lingjian Wang, Guohu Luo, JunFeng Lv 
    Abstract: Electronic information technology plays an important role in intelligent transportation systems, providing efficient, convenient and safe solutions for traffic management. This article explored the application of electronic information technology based on optical sensors in Intelligent transportation systems (ITS) and optimised ITS through electronic information technology. The algorithm used was a transportation path optimisation algorithm based on electronic information technology, which can optimise the transportation path, thereby improving transportation efficiency and reducing transportation costs. Through experiments, it showed that the recognition accuracy of traditional transportation path optimisation algorithms for the shape of goods, colour of goods, length of roads, and length of tunnels was 91.26%, 93.63%, 94.51%, and 92.45%, respectively. However, based on this algorithm, the recognition accuracy for various indicators was 95.47%, 96.22%, 98.91%, and 97.84%, respectively, which indicated that the algorithm proposed in this paper has better recognition performance.
    Keywords: intelligent transportation systems; ITS; electronic information technology; optical sensors; transport experience.
    DOI: 10.1504/IJCSYSE.2025.10060247
  • Security Defence Technology for Webcast Integrating SSA and Reinforcement Learning   Order a copy of this article
    by Delu Wang  
    Abstract: This paper first introduces logistic chaotic mapping and random walk strategy to optimise traditional sparrow search algorithms, and combines them with support vector machines for intrusion detection. Subsequently, reinforcement learning and game model were integrated. The data prove that the loss function of the proposed detection method is the smallest and approaches to 10-6 infinitely when the iteration is 61 times. In the comparison of comprehensive F1 values for detection and defence, when the running time is 0.475 seconds, the F1 value of the proposed method is the highest, reaching 98.31%. In the analysis of defence success rates for different attack strategies, the proposed strategy can achieve a maximum of 99.78% against password intrusion in network live streaming, and can maintain 99.99% against security vulnerabilities in network live streaming security intrusion. This indicates that the proposed security defence technology has implemented various types of network live streaming security intrusion prevention.
    Keywords: sparrow search algorithm; SSA; reinforcement learning; online live streaming; defence; intrusion detection.
    DOI: 10.1504/IJCSYSE.2025.10060356
  • Airborne network security protection technology based on hybrid K-means algorithm   Order a copy of this article
    by Yunna Shao, Bangmeng Xiang 
    Abstract: In order to reduce the security risks such as illegal acquisition of airborne network data and malicious attacks. Based on k-tree structure, weighted density method is used to accelerate K-means clustering. Weighted voting rules are proposed to enhance the training of labelled data sets. Finally, binary tree structure is used to design the classification model. The results showed that the detection rates of remote to local (R2L) and user to root (U2R) were increased by 7.98% and 7.64%, respectively. The research methods achieved 91.63%, 92.29%, 90.68% and 96.34% of the network information confidentiality, integrity and availability, and virus detection ability, respectively. The increases were 36.15%, 40.81%, 44.41% and 44.38%, respectively. The research model can detect airborne network attacks more comprehensively and accurately than the existing detection methods. It can be used to protect the personal information of network users, as well as the security of network communication processes.
    Keywords: K-means algorithm; airborne network security; semi supervised hierarchical classification; tri-means; Kd-tree; detection accuracy.
    DOI: 10.1504/IJCSYSE.2025.10060414
  • A study on the application of 3DHOG-assisted technology in physical education movement recognition   Order a copy of this article
    by Yu He, Na Chen 
    Abstract: An image feature extraction technique based on Histogram of Oriented Gradients (HOG) technology is proposed as a method for human body detection, while 3D Convolutional Neural Networks (3D CNN) technology is combined as a key technology for action recognition, and the two are combined to generate 3DHOG assistive technology applied to the physical education video parsing. The results show that the false recognition rate of the 3D CNN model in the training set is stable around 0.03, corresponding to a Loss of 0.05. The average accuracy of each action of the 3D HOG model is 96.25%, while the recall rate of the model is 81.2%, and the Mean Absolute Error (MAE) value of 1.18% and Root Mean Squared Error (RMSE) value of 0.91%. The 3D HOG model has superior performance and has good application value for action detection and recognition of physical education videos.
    Keywords: Action recognition; HOG; Human detection; Physical education; 3D CNN.
    DOI: 10.1504/IJCSYSE.2025.10060558
  • Research on Innovation and Entrepreneurship Knowledge Management in Higher Vocational Colleges using Big Data Analysis   Order a copy of this article
    by Xiaoyue Xu  
    Abstract: In response to the current lack of a comprehensive innovation and entrepreneurship education system and intelligent evaluation methods in many universities, this study designed a course big data analysis model based on k-means clustering and FP growth algorithm to obtain the degree of correlation between different courses and innovation and entrepreneurship practices. The research results show that the combined algorithm spends less time in mining a large amount of data than apriori derived association rule mining algorithm, FP tree* algorithm and MDML-GA algorithm. And the FP-growth algorithm mining course data found that the results of innovation and entrepreneurship practice are highly correlated with the results of basic theory, with a confidence of 0.91. Therefore, the algorithm proposed in the study has advantages in analysing the influencing factors of students’ innovative thinking and entrepreneurial ability, and is also of great significance in promoting the reform of teaching methods.
    Keywords: educational model; FP growth algorithm; higher education; innovation and entrepreneurship; K-means clustering algorithm.
    DOI: 10.1504/IJCSYSE.2024.10060564
  • The application and analysis of big data technology in the field of smart tourism   Order a copy of this article
    by Mei Xiao, Jiangcen Wang 
    Abstract: As one of the representative technologies of modern digital technology, big data technology plays an important role in data processing, analysis, collection and storage, in the process of tourism development, if we want to realise smart tourism, then the development and utilisation of data information is essential, especially in the use of big data technology to build various smart tourism platforms, the effect of big data technology is very important. In addition, the application of big data technology in the field of smart tourism also includes the provision of accurate and timely tourist information, such as location, number of visitors, ticket purchase, etc. for smart tourism managers and staff, in order to help tourism managers and staff in the decision-making, so as to ensure that tourists get high-quality, high-efficiency, human-oriented tourism services, increase the scenic spots, scenic spots of the tourism value and economic benefits.
    Keywords: big data technology; intelligent tourism; field application; platform analysis.
    DOI: 10.1504/IJCSYSE.2024.10060570
  • k-means on application of means clustering in innovation and entrepreneurship sustainability education in universities   Order a copy of this article
    by Weiyan Chen, Weibo Zheng 
    Abstract: Entrepreneurship education can help improve national competitiveness, and how to achieve sustainable development of entrepreneurship is a difficult problem to be solved. So this study proposes an improved K-means cluster analysis method based on literature and use data analysis to study the sustainable development of entrepreneurship. The improved K-means clustering method is more effective and efficient, with better clustering effect. By using the algorithm to analyse the influence of sustainable institutional environment on college students' entrepreneurship, when considering different institutions, the proportions of those who are willing to choose entrepreneurship are 68.1%, 87.5%, 65.5%, 80.3% and 89.6% respectively. Using the K-means algorithm can accurately reflect the situation of each student, and can grasp the learning needs of different students based on the obtained results. It provides rich and targeted educational resources for each student, providing a good development direction for personalised teaching methods in entrepreneurship education.
    Keywords: k-means clustering; innovation; entrepreneurship; sustainable development; education.
    DOI: 10.1504/IJCSYSE.2025.10060874
  • Bank Marketing Model Based on Improved Neural Network Algorithm   Order a copy of this article
    by Tongdi Hou, Jie Chen 
    Abstract: In commercial banks, traditional marketing methods cannot directly and accurately predict customer needs and preferences, leading to a decline in bank competitiveness. With the progress of big data, deep learning has been applied in many fields. CNN has the characteristics of high-dimensional data and nonlinear data processing. Research using CNN to design marketing models, introducing gravity search algorithm to solve the problem of uncertain network structure selection and overfitting, and using bagging ensemble learning algorithm integration to improve generalisation ability. Due to the uncertainty of network structure in the simulated annealing algorithm, this algorithm was chosen to optimise CNN for comparison. The experiment showed that the CNN MSE optimised by the study was 0.0096, and compared with the comparative model MSE = 0.1021, the similarity between the predicted value and the actual value reached 87%. Therefore, the marketing model based on gravity search algorithm optimisation and bagging integration has good development potential.
    Keywords: Convolutional neural network; GSA; Simulated annealing algorithm; Bagging Integration; MSE evaluation.
    DOI: 10.1504/IJCSYSE.2025.10061137
  • Intelligent Recognition English Translation Model Based on Speech Recognition   Order a copy of this article
    by Xiulian Han, Yawei Ran 
    Abstract: This article uses the physical model sampling survey method, mapping method and parameter analogy method to collect data, analyses the practicality of speech recognition from the four aspects of the model's translation speed, efficiency, language sense and connectivity, and creates a translation model suitable for intelligent recognition. The research results found that in terms of translation speed evaluation, there were 268 samples with the same evaluation by machines and humans, with a consistency rate of 96.58% and a correlation coefficient of 0.74; in terms of language perception evaluation, the consistency rate reached 99.87% and a correlation coefficient of 0.512; in terms of translation efficiency evaluation, the consistency rate was as high as 96.87%, and the correlation coefficient was 0.554; in terms of connectivity evaluation, the consistency rate was as high as 95.19%, and the correlation coefficient was 0.614.
    Keywords: speech recognition; speech signal; English translation model; translation speed.
    DOI: 10.1504/IJCSYSE.2026.10061819
  • Impact of Computer Intelligent Healthcare Combined with Nursing Monitoring on the Efficacy and Medication Safety of Critically Ill Patients   Order a copy of this article
    by Guiqiang Ren, Yuan Zheng 
    Abstract: The article selected 120 severely ill patients admitted to a general hospital in a certain city from January 1, 2022 to December 31, 2022 as the research subjects, divided into an observation group (60 cases) and a control group (60 cases). The observation group used the system as auxiliary treatment, and the control group used traditional methods for treatment. In terms of changes in inflammatory indicators, the P values of reactive protein (CRP), neutrophil percentage (NEUT%), and lymphocyte percentage (LYM%) in the observation group upon admission and 14 days after treatment were 0.001, 0.033, and 0.026, respectively, showed statistically significant differences; the P values of the average changes in CRP, NEUT% and LYM% indicators of patients in the control group at admission and 14 days after treatment were 0.048, 0.206 and 0.118 respectively, and there was no significant difference in NEUT% and LYM%.
    Keywords: efficacy of critically ill patients; drug safety; computer intelligent healthcare; nursing monitoring; medical decision making.
    DOI: 10.1504/IJCSYSE.2026.10061820
  • Abnormal behaviors identification method of college students by WOS-IForest under smart campus   Order a copy of this article
    by Ronghua Teng, Shuyu Teng, Junpeng Wang 
    Abstract: Based on isolated forests, a weighted optimum sub forest algorithm is therefore constructed and examined in response to the tiny fraction of aberrant data and large variations from normal data. Subsequently, a twin gated recurrent neural network model based on linear discriminant analysis loss function is examined and built using the features of data from college students. Ultimately, integrating the two results in a mechanism for recognising aberrant conduct in college students. The research results show that the algorithm proposed in the study has the shortest running time in different dimensional datasets, with an average running time of 124.5 ms and a maximum average accuracy of 98.76%. The average accuracy of the model designed for the study was 98.01%. Finally, the study employed the recognition approach of abnormal behaviour among college students to build a digital image of the students’ aberrant activity, with a pretty broad presentation impact.
    Keywords: smart campus; WOS-IForest; abnormal behaviour identification; twin network; digital portrait.
    DOI: 10.1504/IJCSYSE.2026.10061823
  • Evaluation on Embedded Computer Information Network System Security Architecture   Order a copy of this article
    by Jing Liang  
    Abstract: With the continuous development of internet technology and computer technology, the security performance of information network system has begun to be widely concerned, and an enterprise's computer information network would cause serious consequences if it is hacked. For the security problem of the network system, this paper solved this problem by introducing the embedded system into the information network system. This method is based on the existing computer information system and utilises embedded technology to improve the security performance of the system and prevent advanced trojans from invading the information network. By embedding embedded systems into the existing information network systems of the enterprise, the average integrity increased from 92.8% to 97.3% in two months. For information network systems without embedded systems, the average integrity has dropped from 87.8% to 86.3%, which is already not high and has also decreased. The experimental results indicated that embedded systems are very effective in improving the security performance of computer information network systems.
    Keywords: system security; embedded system; computer information network; trusted platform module.
    DOI: 10.1504/IJCSYSE.2026.10062105
  • Machine Learning Model of English Language Psychology Based on Data Mining Technology   Order a copy of this article
    by Hanhui Li, Zijiang Zhu, Chen Chen, Yi Hu 
    Abstract: The development of the times cannot be separated from language, which is the most basic tool for communication and the carrier for obtaining information. In recent years, data mining technology has achieved relatively good development, especially at the level of students English learning. Therefore, this paper built a machine learning model of English language psychology based on data mining technology. Compared with classroom teaching education, modern distance education was more conducive to the improvement of students' English performance, which could guide students to learn online and offline, and break the space of learning English. Therefore, it is meaningful to study the machine learning model of English language psychology based on data mining technology in this article.
    Keywords: machine learning model; English language; data mining technology; distance education; traditional learning model.
    DOI: 10.1504/IJCSYSE.2026.10062107
  • Application of LightGBM Algorithm in Risk Control of Investment Industry   Order a copy of this article
    by Zhao Guang 
    Abstract: Bond default risk has the potential to result in losses for investors, which might influence their choice of investments. The study uses the gradient boosting decision tree algorithm framework as its starting point, choosing the indicators from the four categories of macro factors, debt characteristics, financial factors, and non-financial factors. The study then further calculates the value of the information in order to screen out the final indicators, and constructs a bond default prediction model. The model is optimised by introducing genetic algorithm to get the final optimised bond default risk warning model. The results of experimental revealed that the model's accuracy was improved by 2.3% in comparison to using a single index factor, the corresponding AUC value after incorporating the study’s proposed index system into the model reached 0.9992, and the standard deviation of the model in the ten-fold cross-validation reached 0.0011. Results indicated that, when compared to the pre-improvement technique, the true rate of the study's improved model was 5.4% higher and the false-positive rate was 0.52% lower. It demonstrates that the model has higher predicted accuracy in addition to superior predictive stability, which can serve as a decision-basis for risk control in the investment industry.
    Keywords: LightGBM; indicator system; risk control; bond default; genetic algorithm.
    DOI: 10.1504/IJCSYSE.2026.10062133
  • Hierarchical Planning and Design of Landscape Architecture Environment Based on VR Technology and Computer Vision Technology   Order a copy of this article
    by Siyi Wang, Ling Wei, Fang Wang, Rongyuan Xiong 
    Abstract: In the traditional environmental level planning and design of landscape architecture, the relevant personnel mostly rely on the analysis and modification of drawings to carry out the project construction and construction plan, and the design results are poor. In order to improve the intuitiveness and authenticity of designers in space vision, this paper analysed the use of virtual reality (VR) technology and computer vision (CV) technology to achieve the design of landscape architecture. This article analyses the terrain, topography, and hydrology in gardens using CV technology, providing a basis for virtual terrain design and hydrological modelling. The results showed that it can be seen that Scheme 2 has high intuitiveness and authenticity. With the continuous progress of science and technology, VR technology would play an increasingly important role in landscape architecture design.
    Keywords: virtual reality; VR; computer vision; CV; landscape architecture; levels of detail.
    DOI: 10.1504/IJCSYSE.2026.10062144
  • Basketball Trajectory Tracking Based on Machine Vision Image   Order a copy of this article
    by Tao Liu, Qinhong He 
    Abstract: Based on the existing experience, this paper made a comparative study of the two existing basketball track tracking methods and technologies, which were background difference method and frame difference method. Two methods were used to compare the same group of video images. The corresponding data were obtained according to the calculation methods of the two methods, and the two groups of data were compared and analysed to show the advantages and disadvantages of the two methods. According to the data comparison of the above two methods, the background difference method had fast calculation speed and high real-time efficiency. The corresponding conclusion was drawn that when high-speed objects needed to be tracked in real-time, the background difference method could be used. When it was not necessary to track the moving track of the target in real-time, the frame difference method could be used to track the moving track.
    Keywords: basketball trajectory tracking; machine vision image; Kalman filtering; image recognition.
    DOI: 10.1504/IJCSYSE.2026.10062145
  • Motion Video Evaluation and 3D Human Motion Simulation in Image Processing Oriented Sports Training   Order a copy of this article
    by Dandan Fan, Xiaodan Yang 
    Abstract: In this paper, image processing technology was used to analyse sports video in sports training, and 3D human motion simulation was carried out. In this paper, firstly, image processing of sports training video was carried out, including image transformation, greyscale transformation and image filtering. After that, the research of moving object detection was carried out, and the detection methods included inter frame difference method and background difference method. The common human models were introduced. Next, a 3D human motion model was constructed, and 3D human knowledge was recognised. In the experiment part, the video of running, walking, rope skipping and sit ups were analysed and simulated. It can help athletes better analyse the mistakes in the sports process, and improve the athletes' sports efficiency, so as to maximise the athletes’ sports ability.
    Keywords: 3D human motion simulation; motion video analysis; physical training; image processing; electronic imaging.

  • Digital Network Communication Strategy of Brand Influence under the Background of Computer Multimedia Technology   Order a copy of this article
    by Xinyi Liu  
    Abstract: Internet has become an important channel for people to get information. Only by timely understanding of Internet trends can enterprises better communicate with consumers and effectively promote and shape their brands. This paper aims to explore the digital network communication strategy of brand influence under the background of computer multimedia technology. This paper analyses the difference between digital technology and traditional technology, probes into the influence and value of digital technology on brand communication, puts forward the image quality evaluation method, and conducts an experimental study on the digital network communication of brand influence. The experimental results show that under the background of computer multimedia technology, the brand awareness score is between 8.5 and 9.3 points, the brand interaction score is between 8.8 and 9.5 points, the brand satisfaction score is between 8.7 and 9.6 points, and the brand loyalty score is between 8.6 and 9.5 points.
    Keywords: brand influence; computer multimedia; digital network; communication strategy; image quality evaluation.
    DOI: 10.1504/IJCSYSE.2026.10062146
  • Computer Vision-based Accurate Identification System for Damaged Parts of Athletes' High-strength Sports Injury Images   Order a copy of this article
    by Guoyang Huang  
    Abstract: With the continuous development of society, the application of computer vision (CV) is also increasing. CV is an important branch of AI. The problem it needs to solve is to understand the content in the image. Due to the fact that various parts of the body would be damaged to different degrees during the high-intensity exercise of sports athletes, the image recognition and analysis must be carried out during the treatment. The accuracy and efficiency of the existing relevant technologies to identify and process them are very low. To solve this problem, this paper proposed a high intensity motion damage image based on fish swarm algorithm, and applied it to gray scale conversion and damage recognition. By comparing particle swarm optimisation (PSO), genetic algorithm (GA) and the algorithm designed in this paper, the experiment in this paper was analysed from two aspects of recognition rate and time. According to the experimental data, when the number of recognised images was 50 and the number of experiments was 50, the recognition rates of PSO, GA and this algorithm were 64.33%, 66.86% and 94.57% respectively. When the number of recognised images was 35, the recognition time of PSO, GA and this algorithm was 0.768 s, 0.807 s and 0.532 s respectively. It was not difficult to see that the design method in this paper had excellent performance in recognition rate and recognition time. Therefore, the system designed in this paper was worthy of further promotion and application.
    Keywords: sports injury; computer vision; CV; sports athletes; accurate identification of image damaged parts; fish swarm algorithm; high-intensity sports.
    DOI: 10.1504/IJCSYSE.2026.10062164
  • Virtual Simulation Technology of Embedded Systems in Multimedia Digital Signal Processing   Order a copy of this article
    by Shujuan Qu  
    Abstract: In order to improve the security of the embedded system, people have connected the embedded system with virtual simulation technology. In embedded systems, this article used virtual simulation technology to analyse embedded systems based on virtual simulation, and completed virtual simulation of reliability enhancement technology. By analysing the solutions of virtualisation technology in multimedia digital signal processing, a virtual simulation signal processing system was studied. Through experimental data, it has been proven that virtual simulation technology had better performance in signal frequency, transmission bandwidth, and signal denoising in embedded systems. The average value of the signal output cut-off frequency of Gaussian white noise was 3% higher than the signal transmission cut-off frequency.
    Keywords: embedded system; field programmable gate array; FPGA; digital signal processing; virtual simulation technology.
    DOI: 10.1504/IJCSYSE.2026.10062165
  • Using Artificial Intelligence to Construct a Character Expression and Action System for a 3D Human Model   Order a copy of this article
    by Bozuo Zhao, Danping Zhan, Canlin Zhang 
    Abstract: In recent years, with the continuous development of computer graphics technology and the wide application of artificial intelligence technology, three-dimensional human modelling technology based on artificial intelligence has gradually become a research hotspot. This article aims to use artificial intelligence to optimise the design of the system. The article introduces common 3D human modelling methods, and then optimises the 3D human reconstruction algorithm. Then, it elaborates on the process of generating complex virtual scenes and 3D facial modelling methods, and uses sequence images to achieve 3D human model reconstruction. Finally, a detailed analysis is conducted on the construction of a character expression action generation system. The experimental results show that the three-dimensional human body reconstruction algorithm designed in this paper reduces the time consumption by about 50% compared to traditional algorithms, and the error is reduced by about 30% compared to traditional algorithms.
    Keywords: virtual reality technology; 3D virtual human; model construction; character expression action system; artificial intelligence.
    DOI: 10.1504/IJCSYSE.2026.10062235
  • Construction of a Network Platform for Student Behavioral Health Monitoring Based on Decision Support   Order a copy of this article
    by Keke Wang  
    Abstract: In the introduction, the significance of research on student behavioural health was introduced, and then academic research and analysis were conducted on the two key sentences of student behavioural health monitoring and decision support in building a monitoring network platform; an algorithm model was established, and decision support algorithm for student behavioural health assessment were proposed to provide theoretical basis for the construction of a network platform for student behavioural health monitoring based on decision support; at the end of the article, a comparative simulation experiment was conducted and the experiment was summarised and discussed; in the last experiment, based on the excellence evaluation criteria, it was calculated that the number of people who evaluated excellent before use was 11% of the total number, while the number of people who evaluated excellent after use was 33% of the total number.
    Keywords: health monitoring; online platform; decision support; student behaviour; health evaluation.
    DOI: 10.1504/IJCSYSE.2026.10062435
  • Optimization Management Method of Enterprise Logistics Supply Chain Based on Artificial Intelligence(AI)   Order a copy of this article
    by Mo Kuang, Lili Weng, Da Kuang 
    Abstract: This article systematically analyses the specific current situation of the entire supply chain using value stream mapping (VSM) tools, and then optimises it from three aspects: real logistics, information flow, and time flow, in order to explore the management efficiency, supply chain costs, and supply chain risks of the logistics supply chain. In order to verify the effectiveness of AI in optimising enterprise logistics management methods, this paper selected the logistics SCM business segments of 12 listed enterprises as the experimental objects for comparison before and after, and evaluated the logistics SCM efficiency, cost management and supply chain risk respectively. The experimental results show that the optimisation management method of enterprise logistics supply chain based on AI had obvious effect on solving the problems existing in enterprise logistics supply chain, and the overall average improvement range was 23.38%.
    Keywords: logistics supply chain; supply chain management; physical distribution management; artificial intelligence; enterprise logistics.
    DOI: 10.1504/IJCSYSE.2026.10062508
  • Sustainable Development of Green Finance in the Low Carbon Economy Era of the Internet of Things   Order a copy of this article
    by Chunshu Wang, Wei Bai, Li Zhao 
    Abstract: This paper used the method of combining theoretical analysis and empirical research, starting from the essence of green finance and low-carbon economy, to explain the necessity of financial institutions to carry out low-carbon finance. Based on this, a green financial development model based on LCE of the internet of things (IoT) was proposed to solve the problem of transformation and upgrading of financial institutions. This paper compared the traditional financial model with the green financial model under the low-carbon background. The results showed that the green financial development model has increased the market size of enterprises by about 4.62%, and the enterprise risk has been effectively controlled, reducing the enterprise operating costs. The vigorous development of green finance can further optimise the industrial structure and improve the allocation of resources, which is of great significance to promote the healthy and stable development of social economy.
    Keywords: green finance; low-carbon economy; LCE; internet of things; IoT; energy report; sustainable development.
    DOI: 10.1504/IJCSYSE.2026.10062591
  • Research on the generation of correlation relations of electricity transmission based on improved Jaro-Winkler algorithm   Order a copy of this article
    by Xiangrui Zong, Bing Feng, Ning Liu, Yuefan Du, Jian Zheng, Bin Zhou 
    Abstract: At present, the data correlation query method in the field of electric power marketing has problems such as low efficiency and low accuracy. This paper improves the Jaro-Winkler character similarity algorithm by combining the editing distance algorithm to improve the matching rate of field names in the data table. Experimental results based on 2,356 data tables show that the improved algorithm is applied to the data table association relationship query, and its accuracy reaches 98%. Based on the improved Jaro-Winkler algorithm and Echarts framework, a visual display system of association relationship of power marketing data table is developed, which provides auxiliary support for business personnel to use data independently and efficiently.
    Keywords: Jaro-Winkler; string similarity; electricity marketing database; associative relationships.
    DOI: 10.1504/IJCSYSE.2026.10062594
  • Optimisation Evaluation of Middle and Bottom Level Scheduling Algorithms Based on Embedded Wireless Communication and Big Data Query Processing Technology   Order a copy of this article
    by Haifeng Chen, Lili Ding 
    Abstract: In embedded wireless communication system and big data query processing technology, the quality of task scheduling algorithm largely determines the performance of the system, and how to optimise the real-time scheduling is a problem worth studying. In this paper, the scheduling algorithm analysis and big data query processing technology of embedded wireless communication system are optimised, the specific algorithm optimisation steps are given, and the system structure design diagram and operation process are analysed. This paper takes the classical elevator scheduling problem as the research object, studies the process correlation and response time of three scheduling algorithms, FCFS, RR and PSA, and uses the benefit function R in the optimisation content of the algorithm to analyse the values of the scheduling algorithm nos. 18 when R = 1 and R = 2. Research shows that most scheduling algorithms only consider queuing order and have randomness in the local distribution of data.
    Keywords: scheduling algorithm; wireless communication; embedded system; big data; query processing; elevator scheduling.
    DOI: 10.1504/IJCSYSE.2026.10062938
  • E-commerce Customer Marketing Classification Technology Based on The Improved Ant Colony Clustering Algorithm   Order a copy of this article
    by Ming Zhong  
    Abstract: The vigorous development of the internet has driven the development of the e-commerce market economy. Facing a huge number of consumers, every e-commerce enterprise is facing the problem of customer classification. To address this issue, the collected customer characteristic data are processed by feature selection and principal component analysis for dimensionality reduction. According to the ant colony clustering model, a new two-dimensional data object load state matrix is introduced, and by improving the ant's. Observe the radius and introduce the Sigmoid function to improve the test accuracy. Test findings demonstrate that the F-measure value of the standard model is 0.846, and the F-measure value of the improved model is 0.934. The former has an error of 0.25 after 500 iterations, and the error of the later is 0.12 after 300 iterations. The average consumption time of the standard model test dataset is 51.64 s, and the average consumption time of the improved model is 28.12 s. The test's findings reveal that the improved method has smaller error value and shorter time consumption when dealing with discrete data, and its performance is better than the standard model, which can better classify customers. The growth of e-commerce has been greatly influenced by the research findings.
    Keywords: E-commerce; customer classification; marketing; data processing; ant colony clustering; ACC.
    DOI: 10.1504/IJCSYSE.2026.10062984
  • Research on user behavior detection algorithm of e-commerce platform based on machine learning   Order a copy of this article
    by Yuanyuan Tang  
    Abstract: This paper first introduced ML, including C4.5 algorithm and support vector machine algorithm in decision tree algorithm, and introduced random forest algorithm based on ML. Then, the user behaviour of EC platform was analysed and detected. First, the problems to be solved in the EC platform behaviour analysis are determined. Then, the data was collected, and then the collected data is characterised and analysed. The extracted data was divided into training set and test set, and the algorithm model was used to analyse the data. In the experiment part, three ML algorithms, C4.5 algorithm, support vector machine algorithm and random forest algorithm, were used for data analysis. The performance of user data analysis of the three algorithms was analysed by training set and ten fold cross validation. The relative error of model classification was the lowest, which showed that ML algorithm has good data analysis ability and good application effect in the field of EC platform user behaviour detection.
    Keywords: user behaviour detection; e-commerce platforms; machine learning; artificial intelligence.
    DOI: 10.1504/IJCSYSE.2026.10062985
  • Financial Risk Monitoring and Prevention of Exhibition Enterprises Based on Security Cloud and Edge Computing Framework   Order a copy of this article
    by Jiang Wang  
    Abstract: Therefore, this paper has analysed the financial risk characteristics and causes of exhibition enterprises, and then used security cloud and edge computing to build financial risk monitoring and prevention measures, so as to improve the quality of financial risk management of exhibition enterprises. The financial risk monitoring and prevention effect of exhibition enterprises under the security cloud was higher than that of the original financial risk monitoring and prevention. Among them, the financial risk monitoring effect was 9.4% higher than the original one, and the financial risk prevention effect was 9.7% higher than the original one. In short, both AI and edge computing can improve the financial risk bearing capacity of enterprises.
    Keywords: financial risk monitoring; security element and edge computing; artificial intelligence; AI; network security.
    DOI: 10.1504/IJCSYSE.2026.10063124
  • Design of athlete physical training system based on a smart wearable device   Order a copy of this article
    by Shikai Cai  
    Abstract: Smart wearables are any object that incorporates electronic technology or a gadget worn close to the body. Information may be tracked in real-time with the help of these athletes' progress; coaches no longer need to depend just on timings and splits on precise measurements of position, distance, velocity, and acceleration. The challenging characteristic of such physical training is the athlete's irregular optimality, scalability and generalisability. The gathering and quality of data is a significant obstacle to sports analytics. Even though there is a mountain of data, gathering and cleansing it is not always easy. Data quality is another potential issue; incomplete or erroneous data is a real possibility. Hence, in this research, smart sensors enabled intelligent physical monitoring systems on IoT platform (SS-IoT) technologies, which have been improved for sports monitoring systems with the athlete's physical training. The BP neural network establishes the athletes' physical training for data processing and monitoring in that physical control mechanism. Accurately predicting an athlete's physical state via simulation is a cutting-edge scientific method for increasing the efficiency of physical training. The experimental results show the SS-IoT achieves an accuracy ratio of 90%, efficiency ratio of 90.6%, prediction ratio of 91%, performance ratio of 95%, and error rate of 8.56% compared to other methods.
    Keywords: physical training; athlete; smart sensor; internet of things; BP neural network; data processing.
    DOI: 10.1504/IJCSYSE.2026.10063125
  • Application of Matrix-based Genetic Algorithm in Foreign Trade CRM   Order a copy of this article
    by Yu Tang, You Zhou 
    Abstract: Genetic algorithm is a bionic optimisation algorithm in a macroscopic sense, which can optimise the combination of numerical values. On this basis, this paper proposes a matrix genetic algorithm combined with matrix algorithm, and applies the algorithm to foreign trade CRM, aiming to improve the efficiency of customer relationship management (CRM). This article first introduces genetic algorithm, then analyses the operating mechanism and overall architecture of the foreign trade customer management system. Finally, a mathematical model is established using matrix genetic algorithm, and dynamic management of foreign trade customer relationships is achieved through fuzzy management matrix. Research shows that the algorithm can be used to search for customer information of foreign trade enterprises, extract useful information for decision-making, and the information extraction efficiency is 5.8% higher, so that foreign trade enterprises are in a favourable position in the fierce market competition. It ultimately gets the highest profit.
    Keywords: matrix algorithm; genetic algorithm; foreign trade CRM; matrix genetic algorithm.
    DOI: 10.1504/IJCSYSE.2026.10063347
  • Improve Text Classification Accuracy by Using Fuzzy-Convolutional Neural Network Model   Order a copy of this article
    by Xuan Wang, Jing Su 
    Abstract: Most of the text data we can see in daily life is fuzzy, which fuzzy information will increase the noise and reduce the classification accuracy. In order to solve this situation, this paper proposes a network model that fuses fuzzy neural network (FNN) and convolutional neural network in text classification (TextCNN), namely text fuzzy-convolutional neural network (TextFCNN). Firstly, the model uses FNN and TextCNN to obtain two sets of classification results; secondly, the fuzzy inference system is combined to further eliminate the fuzzy characteristics and achieve more correct classification outcomes. In the movie review (MR) dataset, the model was improved by 1.41% and 6.38% in accuracy compared to the single neural network FNN and TextCNN, respectively. Compared with other text classification methods, the accuracy of TextFCNN is improved by 0.33% 3.19%. Experimental results show that the network model TextFCNN can indeed improve the effect of the classifier.
    Keywords: text classification; fuzzy neural network; FNN; convolutional neural network; fuzzy theory; natural language processing; NLP.
    DOI: 10.1504/IJCSYSE.2027.10063460
  • Social network user browsing trajectory detection based on soft computing to promote a healthy environment   Order a copy of this article
    by Qing Cai  
    Abstract: In order to improve the detection ability of browsing trajectory data for social network users, a mobile computing based method for detecting browsing trajectories of social network users is proposed. The study first utilises fuzzy logic to establish a social network user browsing trajectory data detection model. Then, the fuzzy parameter recognition method is used to extract the features of the browsing trajectory data of social network users. Finally, a social network user browsing trajectory detection method was designed by combining random forest learning algorithm and matched filtering detection method. The experimental results show that the method has a good output signal-to-noise ratio to eliminate redundancy, with a maximum redundancy elimination of 23.7 dB. The accuracy and stability are high, up to 93%, and it has a good detection effect on the browsing trajectory of social network users.
    Keywords: soft computing; social networks; trajectory similarity; browse track; random forest; environment; social media.
    DOI: 10.1504/IJCSYSE.2026.10063532
  • Particle Swarm Optimisation (PSO)-based self-efficacy model for student learning and decision-making capabilities   Order a copy of this article
    by Qing Zhou  
    Abstract: Improving the models structural validity and reliability requires taking into account the students implicit relationship with the decision-making process about their professions. With a self-efficacy model based on social cognitive theory, this article aims to help students interested in education make more informed career selections. This study aims to evaluate the social cognitive theory-based self-efficacy model in order to find its distinctive elements. If you want to help your pupils find a job that fits their hidden talents, you may utilise their implicit feature matrix. When trainees make professional decisions, a supplemental matrix is used to investigate the hidden relationships among them. Compared to its alternatives which relied on cluster analysis and the user portrait method this model exhibited superior structural validity and dependability. The successful model validation provided evidence of this. Therefore, it is a reliable measure of students confidence in their ability to make good career decisions down the road.
    Keywords: social cognitive theory; students; self-efficacy; decision making; particle swarm optimisation; PSO.
    DOI: 10.1504/IJCSYSE.2027.10063633
  • Design of Network Education and Teaching Cooperation Platform Based on Moodle   Order a copy of this article
    by Haimiao Su  
    Abstract: Aiming at the problem of mismatching teaching content in current online learning, this paper studies the construction of online education and teaching collaboration platform based on Moodle platform, and proposes the functional design of teaching platform with collaborative filtering algorithm as the core. The effectiveness of the collaborative filtering algorithm in the platform is evaluated through comparative analysis, and the platform is evaluated by analysing the functional performance effect of the platform in the application. The results show that the Moodle-based teaching platform has a maximum response time of only 0.12 seconds for students' learning resources. The above results indicate that the teaching platform designed by the research can quickly achieve resource push and has significant value in campus course teaching assistance.
    Keywords: network big data; teaching platform; Moodle; collaborative filtering.
    DOI: 10.1504/IJCSYSE.2024.10063793
  • Research on the application of flipped classroom model in college physical education courses in the new media era   Order a copy of this article
    by Feng Ruan, Yajun Cao 
    Abstract: To solve the inefficient teaching in college physical education courses, the study integrates the flipped classroom model into college physical education courses. And the study builds an evaluation model with the BP algorithm improved by adaptive GA and entropy value method. And then the teaching effect of the flipped classroom model of college physical education proposed by the study is evaluated. The performance comparison experiment of the improved algorithm found that the PA of the improved AGA-BP algorithm was 0.97, higher than the other two comparison models. This result indicated that the AGA-BP algorithm could improve the evaluation accuracy of the teaching evaluation model. An empirical analysis of the flipped classroom model of college physical education finds that the flipped classroom model of college physical education proposed in the study can improve students’ physical education performance. It also can effectively enhance students’ positive subjective emotions of learning. The reform of college physical education courses can help college students improve the overall learning of physical education courses and enhance physical quality.
    Keywords: flipped classroom model; college physical education; entropy value method; GA; BP.
    DOI: 10.1504/IJCSYSE.2024.10063795
  • Text classification and topic mining of intelligent forum comments in university MOOC based on CNN networks   Order a copy of this article
    by Xue'e Zhong  
    Abstract: This study conducted an intelligent classification of comments on MOOC forums, categorising texts based on urgency and subject. In emergency level classification, LSTM recurrent neural network is used, and after identifying emergency comments, Bayesian subject mining models and CNN networks are used to perform secondary text classification. In the emergency classification simulation experiment, the RNN-LSTM model proposed in this study has three evaluation indicators: overall accuracy, recall rate, and F1 value, which are 0.91, 0.949, and 0.94, respectively, which are higher than conventional classification methods. In the practical application scenario of text mining, the probability of misjudgement of Bayesian network combined with CNN model is 4%. Research has shown that the MOOC forum intelligent management comment classification method proposed in this study can effectively increase the feedback efficiency of teachers in MOOC forum comment management, and improve the feedback communication effect of teacher-student education communication.
    Keywords: recurrent neural network; RNN; LSTM; Bayesian networks; CNN; Catechism forum.
    DOI: 10.1504/IJCSYSE.2025.10063796
  • A visual presentation of English online teaching information from a digital perspective   Order a copy of this article
    by Feifei Dang  
    Abstract: English online teaching needs auxiliary information when displaying teaching contents and methods. How to visually display this information has become the research focus. In this study, the cyclic neural network is introduced to extract the features of text and image. To solve the short storage time in the cyclic neural network, the long- and short-time memory network and gating unit are introduced. In addition, the stacking attention mechanism is introduced to improve the accuracy of text and image feature extraction. The results show that, in the datasets Flickr-30K and MS-COCO, the recall and accuracy of the new model are higher. When dealing with textual data, its A, R and F values are 0.892, 0.876 and 0.883, respectively; its maximum accuracy is 93.57%. It indicates that the attention mechanism can effectively improve the algorithm performance. The visualisation method based on neural network improves the display effect of English online teaching information.
    Keywords: digital; English; online teaching; information visualisation; presentation methods.
    DOI: 10.1504/IJCSYSE.2024.10063803
  • Research on Innovation and Entrepreneurship Education Teaching Method Innovation of College Students Based on Fuzzy Analytic Hierarchy Process   Order a copy of this article
    by Biying Zhuang  
    Abstract: In the era of rapid economic development, IAE has become a vital force and energy for social and economic growth. As the driving force behind the societal development of the future, college students are the main group participating in IAE. Therefore, doing a good job in IAEE for college students is extremely important to social development. In the innovation and development of IAE teaching for college students, most of the teaching effect still adopts traditional evaluation methods, which cannot objectively measure it. Therefore, to better realise the IAE teaching methods development, IAE teaching evaluation indicators based on FAHP is constructed. On the basis of this index system, a TEEM based on RBFNN is designed. Aiming at the poor convergence of RBF model, the RBF model is optimised by the LM algorithm. The findings indicate that the convergence speed of the improved LM-RBF TEEM has been significantly improved. The accuracy rate of the evaluation reaches 98.69 %, which is 1.46 % higher than that of the RBF model. Therefore, the teaching effect assessment model based on the improved RBFNN can better evaluate the teaching effect of IAE, and realise the innovative development of IAE teaching methods.
    Keywords: fuzzy analytic hierarchy process; innovation and entrepreneurship education; teaching method; RBF; LM algorithm.
    DOI: 10.1504/IJCSYSE.2025.10063884
  • A data mining-based model for comprehensive assessment of English teaching quality of university students   Order a copy of this article
    by Yanchen Li 
    Abstract: The purpose of this thesis is to explore the effectiveness of English classroom practices in colleges and universities, and to construct a complete teaching quality assessment model using data mining methods. The assessment model consists of two parts, the first of which is the education of students. The second section is learning. The experimental results showed that this model had a correctness of 0.985 and had a recall of 0.982. The sum of the least errors was 726. Feed-forward neural network was a forecasting method that was applied to online learning materials and compared with the traditional linear regression method. The method was found to have higher accuracy with a prediction accuracy of 0.974. The experiment showed that the established method for assessing English language teachings effect is feasible and effective.
    Keywords: K-modes algorithm; co-occurrence; feed-forward neural network; teaching quality.
    DOI: 10.1504/IJCSYSE.2024.10063885
  • Application of Multimedia Interaction in Museum Display Space Design   Order a copy of this article
    by Bin Wang 
    Abstract: The visual based gesture recognition algorithm was used to optimize the museum interactive project. Through the scientific control, the visitors' scores before and after the project optimisation were investigated and counted. The score range was 0-100. The higher the score, the better the experience. The statistical results showed that before using the algorithm to optimize museum interactive projects, the average scores of visitors on virtual books and interactive projection were 85.6 and 89.03, respectively. After optimisation, the average scores for these two projects were 93.51 and 95.42, respectively. Based on this, it could be concluded that the optimised interaction method had higher attractiveness and fun and could better attract the attention of visitors and enable them to have a deeper understanding of museum exhibits and knowledge. This further proved the effectiveness and importance of vision based gesture recognition algorithm in museum interactive projects.
    Keywords: Museum Display Space; Interactive Mode; Gesture Recognition Algorithm; Virtual Book.
    DOI: 10.1504/IJCSYSE.2026.10063886
  • Data Collection and Protection of Personnel Evaluation under Differential Privacy   Order a copy of this article
    by Yue Wu, Yaping Pan, Gang Wang, Shenghong Wang, Zhenfen Zhang 
    Abstract: This article utilized DP (differential privacy) technology to ensure that sensitive information of individuals was effectively protected during data collection and processing. Firstly, it used the LN (Laplace Noise) distribution to perform DP protection on the raw data, and compared it with exponential noise and Gaussian noise. Secondly, it divided the data into unrelated groups, added noise to each group, and adjusted DP parameters to balance data protection and data availability. Then, this article utilized the GD (Gradient Descent) algorithm to optimize parameters to maximise data availability and allocate privacy budgets to different data processing operations. Finally, during the data collection process, this article randomly selected samples and introduced random trap data points to reduce the risk of individual identification. It used metrics such as information loss and KL (Kullback Leibler) divergence to evaluate the degree of privacy leakage in DP protection.
    Keywords: Data Privacy Protection; Differential Privacy Techniques; Information Noise Addition; Privacy Measurement Metrics; Individual Identification Risk.
    DOI: 10.1504/IJCSYSE.2027.10064042
  • Personnel Evaluation of Data Encryption Transmission and Storage Technology for Cloud Computing Environment   Order a copy of this article
    by Xu Zhang, Peidong Du, Qingzhao Hu, Zuohu Chen, Miao Wang, Long Wang 
    Abstract: In response to the current evaluation by personnel that data encryption transmission and storage technology has problems such as limited storage space resources, slow encryption speed, and low encryption accuracy. This article studied the existing problems in the cloud computing environment. In the cloud computing environment, stream cipher encryption methods were utilised to encrypt movie evaluation data, and combined with chaotic sequence systems, the encryption and decryption process of the data was completed. Then, transport layer security protocols and hash functions were utilised to verify user data and ensure data integrity. At the same time, data storage technology in cloud computing systems was studied, and unstructured storage technology was utilised to store evaluation data, effectively improving the speed of data storage. The encryption accuracy of data transmission obtained by the stream cipher encryption method was above 96.85%, and the average encryption accuracy of 50 experiments was 97.95%, which was 9.01% higher than the average encryption accuracy of the digital signature algorithm method. In response to the problem of limited storage space resources in cloud computing environments, this article utilised stream cryptography and unstructured storage technology to effectively ensure the security of data transmission and improve the space and speed of data storage.
    Keywords: unstructured storage technologies; USTs; evaluation of data; encrypted transmission; cloud computing environments; stream cipher; homomorphic encryption.
    DOI: 10.1504/IJCSYSE.2027.10064044
  • RBF Neural Network Model Construction for Enterprise Financial Big Data Analysis   Order a copy of this article
    by Na Feng 
    Abstract: The study builds a system of financial indicators first, and then uses the fast density peak clustering (FDPC) algorithm and the Adam algorithm to optimise the radial basis function (RBF) network to create a model for predicting financial risk. The results reveal that the initial accuracy of the FDPC Adam RBF model is higher than 60%, and it tends to converge at four iterations, resulting in an accuracy of 95.6%. The FDPC Adam RBF model achieved a minimum value of 0.183 in mean square error (MSE). In summary, it can be seen that the RBF neural network model for enterprise financial big data analysis is significantly better than other common neural network models in terms of computational efficiency and prediction accuracy, making it more suitable for deep analysis of financial data and risk warning. This conclusion provides strong support for the application of advanced artificial intelligence technology in the financial field.
    Keywords: financial crisis; financial indicators; radial basis function; RBF; fast density peak clustering; FDPC; Adam.
    DOI: 10.1504/IJCSYSE.2027.10064058
  • The Regulation Method of Agricultural Internet of Things Services Based on Dynamic multi-objective Optimization   Order a copy of this article
    by Chaoqun Huang, Qianlan Liu, Wenbin Qian 
    Abstract: In response to the complex and ever-changing environmental impacts faced in the current construction of agricultural internet of things technology. A supervision method for agricultural internet of things services based on dynamic multi-objective optimisation is proposed. The poor dynamic capabilities in the intelligent agricultural internet of things can be solved by constructing a decomposed algorithm. According to the findings, it performed well in convergence, hypervolume value and extreme point accuracy. This algorithm could propose the optimal service matching scheme based on a single-target service strategy, with good diversity. In addition, the calculation time of this algorithm was relatively short. Compared with the other two comparison methods, it led by 3.59s and 8.39s, respectively. Meanwhile, the average service cost of this algorithm was relatively low. It reduced the average service cost by 16.39% and 25.00%, respectively. Overall, the dynamic multi-objective optimisation agricultural internet of things regulation method has performed well in practical application, significantly improving accuracy. It can provide the highest quality service at the lowest cost within the shortest service time. In summary, this research provides an effective solution for the regulation of internet of things services in the intelligent agriculture.
    Keywords: internet of things; IoT; MOO algorithm; dynamic multi-objective optimisation algorithm; agriculture.
    DOI: 10.1504/IJCSYSE.2027.10064212
  • Application of MOOC+SPOC Mixed teaching in Athletics Professional Courses in colleges and universities   Order a copy of this article
    by Xiaoqin Guo  
    Abstract: In the context of Education Modernisation 2.0, this paper discusses the current situation of athletics teaching in colleges and universities, and expounds the application of MOOC+SPOC mixed teaching in athletics teaching in colleges and universities. In the post-COVID-19 era, MOOC+SPOC combined teaching has become an important orientation in the teaching reform and innovation of many universities in China. Many practical activities show that the mixed teaching method has promoted the reform and innovation of teaching, and has achieved certain results in improving the actual teaching effect. MOOC are used to fill the defects of SPOC teaching network resources, to carry out purpose-oriented teaching with SPOC, to deal with many problems caused by the wide coverage of MOOC and the lack of constraints caused by the wide coverage, and to carry out mixed teaching combined with physics courses, to complete the co-creation of disciplines in ordinary high school athletics teaching.
    Keywords: massive open online courses; MOOC; small private online courses; SPOC; mixed culture education; universities; athletics.
    DOI: 10.1504/IJCSYSE.2025.10064269
  • A Network Model for a Mobile Learning Environment to Track Students' Progress   Order a copy of this article
    by Chengliang Huang, Fumin Zhang, Xiaotong Li 
    Abstract: In this paper, we present a model for evaluating the education that is based on a network and an approach to its construction and assessment. The educational model that exhibits quality evaluation has better practicability during common sense application and fully satisfies the test requirements, according to studies, when seen through the lens of school effort partnership. The degree of information that both the students and the teachers possess is a significant in determining the instructional strategies that are utilised for both learning and teaching. The primary focus of attention is directed toward the aspects of the classroom environment that have been recognised as being disruptive. This research has a primary purpose of identifying these elements and offering ways for successfully managing or removing them in order to improve mathematics teaching and learning.
    Keywords: Learning capacities of students; model for evaluating networks; cloud computing; mobile learning environments.
    DOI: 10.1504/IJCSYSE.2027.10064743