International Journal of Computational Systems Engineering (24 papers in press)

Regular Issues

• A data mining-based approach to integrating multimedia English teaching resources  
  by Ran Li 
  Abstract: With the rapid development and popularisation of information technology, great changes have taken place in the field of education. Multimedia technology has gradually become an important means of English teaching. However, the integration of multimedia English teaching resources needs to be improved. Therefore, the D-K-means algorithm is formed by adding the splitting and aggregation operations to the clustering process of K-means algorithm to cluster the teaching resource data, and the apriori algorithm is used to find the valuable association rules in the data. Finally, the simulation experiment is carried out. The results show that the objective function value of the final solution of D-K-means algorithm is 108.64, which has stronger search ability. After combining with apriori algorithm, the accuracy of data mining can reach about 96%. In practical application, the teaching resources after the integration of this method have significantly improved students’ English abilities, which show that this method can effectively tap and integrate English teaching resources, and provides a realisable path to improve the quality of English teaching.
  Keywords: data mining; K-means; English language teaching resources; association rules; apriori.
  DOI: 10.1504/IJCSYSE.2022.10053047
   
  
• Business model innovation and development path selection of international cultural trade under circular economy  
  by Yuefeng Han, Guan Wang 
  Abstract: With economy’s rapid development, people have higher requirements for quality of cultural products and services, which requires model innovation and path selection. To study the innovation of international cultural trade business model and the choice of development path, this study first analyses the integration of cultural tourism industry in province A by using the fusion measurement method, then studies the influencing factors by using the grey correlation and regression analysis. From 2012 to 2022, the highest value of cultural industry to tourism industry is 0.052, 0.039, 0.035 and 0.028, respectively; the highest values of tourism industry to cultural industry are 0.043, 0.031, 0.054 and 0.071, respectively, which do not reach 0.1; integration degree between industries increased from 0.0035 in 2012 to 0.0186 in 2017, but decreased to 0.0130 in 2022. Indicating that integration degree between the two industries is low, and the integration development is not stable.
  Keywords: business model; circular economy; cultural tourism industry; development path; international cultural trade.
  DOI: 10.1504/IJCSYSE.2022.10053096
   
  
• Research on an online teaching system for ethnic music courses incorporating fuzzy control and CRP algorithms  
  by Zhihui Yu, Yifan Zhang 
  Abstract: With the continuous development of intelligent teaching, ethnic music teaching in colleges and universities began to try to use the online and offline teaching mode. However, with the continuous improvement of teaching requirements, the traditional teaching system gradually appeared an unsatisfactory state in the use of resources. Therefore, how to effectively improve the use of online teaching resources is the key to improve teaching quality. The emergence of cloud computing has solved the current problem of wasted network resources. In order to achieve resource optimisation and stability for online teaching of ethnic music courses, the study proposes a design scheme for an online teaching system incorporating fuzzy control and CRP algorithms. Firstly, the CRP algorithm is used to achieve resource scheduling, secondly, adaptive fuzzy control is used to enhance computational stability by, and finally, the two algorithms are combined to build an online teaching platform for ethnic music courses.
  Keywords: adaptive fuzzy control; CRP; ethnic music; online teaching.
  DOI: 10.1504/IJCSYSE.2022.10053106
   
  
• A method of forecasting cross-bordere-commerce stocking for SMEs based on demand characteristics and sequence trends under sustainable development strategy  
  by Hua Yang, Lihui Yu 
  Abstract: With the continuous acceleration of economic globalization, cross-border e-commerce enterprises have started to apply big data technology to find business information, among which the accurate forecast of stock availability has become an important influencing factor on consumers’ online shopping experience In order to improve the accuracy of cross-border e-commerce stocking prediction, this study first analyzes the demand feature-based selection and prediction method, followed by the analysis of the serial trend-based stocking prediction method, and then proposes a stocking prediction method based on demand features fused with serial trend, and finally analyzes the results of cross-border e-commerce stocking prediction for SMEs by the proposed method The results show that the contribution rate of Class A goods is the highest, which can be considered to build a stocking warehouse overseas for stocking, while stocking at the origin, and using multiple batches of small lot stocking to reduce inventory costs while ensuring capital
  Keywords: sustainable development; big data; demand characteristics; sequence trends; cross-border e-commerce; stocking forecast.
  DOI: 10.1504/IJCSYSE.2022.10054023
   
  
• Designing remote sharing system of network education resources for software engineering specialty based on web technology  
  by Qin Yang 
  Abstract: With the rise of online education, the traditional offline education model has been greatly challenged, and people gradually tend to choose online education which is more convenient for acquiring knowledge. In order to make full use of online education resources, this study designed a web-based education resource remote sharing system and tested the system using software engineering students. The study first proposed a method design for the remote sharing system of educational resources based on webRTC technology. Based on this, a load balancing method design based on an improved consistency hashing algorithm was proposed considering the problem of overloaded servers, and finally the constructed remote sharing system was tested with the improved algorithm. The experimental results show that the functions of all the modules of the network education resources remote sharing system constructed in this research can be used normally;
  Keywords: Network education; Hash algorithm; Remote sharing system; Web.
  DOI: 10.1504/IJCSYSE.2022.10054098
   
  
• A corpus-based study on the characteristics of the use of spoken English chunks  
  by Rong Hu 
  Abstract: This study constructs the English-speaking SELL corpus, proposes a CNN-LSTM-SA algorithm model-based English speaking recognition technique for the use of English-speaking blocks, and analyses the results of the SELL corpus and the speaking recognition model. The results show that the model’s loss rate shows a trend of slow increase after a sharp decrease. When the number of iterations of the model is 300, the inflection points of the loss value and accuracy rate occur. At this point, the accuracy tends to converge, and its training accuracy is close to 88%, which is significantly higher than other algorithms. The CNN-LSTM network performs the best under the ReLu and tanh functions selected for the study, and the MAE and RMSE indexes 26.54 and 36.11, respectively. The model performance is higher than other algorithms under all six complexities, and its difference is about 4% at the lowest, with a very stable performance advantage.
  Keywords: corpus; spoken English; chunk features; self-attentive mechanism; CNN-LSTM.
  DOI: 10.1504/IJCSYSE.2023.10054652
   
  
• Research on the integration of English online teaching resources based on improved association rule algorithm  
  by Hao Liu 
  Abstract: With a series of achievements of intelligent algorithms in various fields, people began to try to apply intelligent algorithms to information-based teaching and learning, and use relevant data analysis techniques to improve existing teaching models. The study integrates RBF neural networks with association rule algorithms, and then constructs an English web-based teaching prediction model. Using a crawler data collection tool, the English test scores of a university were selected to test the model. The results show that the improved model has shorter running time and can be iterated to a stable state faster. The model was used for the prediction of actual grades, and the average accuracy of the model was obtained as 95.7%. Comparing the relative error values of the prediction model with different influencing factors, we found that the average relative error was only 0.041. The improved model can achieve better results when used for English score prediction.
  Keywords: prediction models; neural networks; association rules; relative error values.
  DOI: 10.1504/IJCSYSE.2023.10054701
   
  
• Exploring the costing method of steel enterprises based on PSO algorithm under the concept of sustainable development  
  by Jinping Qiu 
  Abstract: According to the established raw material cost difference accounting model of iron and steel enterprises, AAD-MOPSO algorithm is used to solve the model. The model, a very accurate cost accounting technique, balances the quantity and price variations of steel goods. The experiments are compared with particle swarm optimisation (PSO) and multi-objective particle swarm optimisation (MOPSO) to verify the excellent performance of the proposed model. The results of AAD-MOPSO algorithm in the SP test function show that its SP, IGD and IH indicators are lower than the comparison algorithm. The best performances of the three indicators are 3.419E-4, 2.154E-4 and 1.017E-3. As a result, the AAD-MOPSO algorithm enhances the iron and steel industry’s cost accounting accuracy and promotes the long-term growth of businesses.
  Keywords: particle swarm optimisation; PSO; costing; sustainability; least squares; multi-objective particle swarm optimisation; MOPSO.
  DOI: 10.1504/IJCSYSE.2023.10054888
   
  
• Visualization and analysis method of enterprise financial expenditure data based on historical database  
  by Ping Chen 
  Abstract: Under the influence of the trend of data informatisation, the storage and utilisation of enterprise financial expenditure information is more dependent on technologies such as databases. How to use enterprise information databases more effectively and make the storage and utilisation of enterprise financial expenditure data more efficient and easier is a concern of users. The study proposes a visual analysis model of enterprise financial expenditure data based on real-time historical database, which is constructed based on auto-encoder and K-mean clustering algorithm. It improves both algorithms in the design process to reduce the negative impact of their defects on the visual analysis model. The performance test of the visual analysis model of financial expenditure data shows that the loss value is as low as 0.02 and the error sum of squares is as low as 0.18. This indicates the value of the model for visual analysis of financial expenditure data of large enterprises.
  Keywords: historical database; financial data; visual analysis; neural network; K-means clustering.
  DOI: 10.1504/IJCSYSE.2023.10055104
   
  
• Influence of Technology Optimization Based on Machine Learning Algorithm on Financial Management Innovation of E-commerce Enterprises  
  by Rui Min 
  Abstract: In order to improve the financial ability of e-commerce enterprises to deal with risks and optimise their financial early warning effect, a complete random forest-based financial early warning method for e-commerce enterprises based on k-nearest neighbours is proposed. Firstly, in order to improve the classification effect of complete random forest algorithm on dynamic data, a complete random forest algorithm based on k-nearest neighbour is proposed; then, on this basis, the financial risk evaluation system of e-commerce enterprises is established by using the analytic hierarchy process, so as to complete the construction of the financial early warning model of e-commerce enterprises, and finally its application effect is tested and analysed. The results show that the minimum prediction accuracy and F1 value of the model remain at 0.7, which are 0.58 and 0.3 higher than the NB model, respectively.
  Keywords: k-nearest neighbour; KNN; random forest; machine learning; e-commerce; corporate finance.
  DOI: 10.1504/IJCSYSE.2023.10055105
   
  
• A transfer learning-basd model for assessing university students’ innovation and entrepreneurship  
  by Yiping Zheng 
  Abstract: In the current trend of innovation and entrepreneurship, the number of students who start their own innovation and entrepreneurship is increasing. In order to enable university students to assess their own innovation and entrepreneurship ability and avoid greater risks, the study starts with a convolutional neural network (CNN) based on migration learning, which is used to establish an assessment model of innovation and entrepreneurship ability. The model is based on migration learning and adversarial networks to strengthen the learning ability and incorporate game theory, with similar adversarial learning of classifiers and generators in the migration, so that the model has stronger learning ability and faster computing speed, and avoids problems such as over-convergence and excessive degrees of freedom. A variational self-encoder is used on the encoder to further improve the accuracy and precision of the input data recognition by compressing the information bottleneck.
  Keywords: transfer learning; convolutional neural network; CNN; evaluation model; innovation and entrepreneurship.
  DOI: 10.1504/IJCSYSE.2022.10055142
   
  
• Data mining research on sustainable business model innovation of enterprises based on particle swarm algorithm  
  by Jianxiong Hu 
  Abstract: To achieve fast and accurate data mining, this research proposes a data mining method based on particle swarm optimisation algorithm, which first introduces a time factor to optimise the fractional-order particle swarm algorithm (TFFV-PSO), and then implements automatic clustering on the basis of improved fractional-order PSO. In the result part, when searching in the early stage, the TFFV-PSO proposed in this paper can avoid forming local optimisation in multimodal function, and has better robustness; the convergence speed of the TFFV-PSO algorithm is faster and more accurate in the two-dimensional case than in the ten-dimensional case; the number of clusters of the new algorithm in different datasets is consistent with the actual, and the correct rates in dataset 1 and dataset 2 can reach the algorithm’s average DBI values are lower. The average DBI values of the algorithm are lower than those of other algorithms.
  Keywords: particle swarm algorithm; enterprise; sustainability; business model creation; data mining.
  DOI: 10.1504/IJCSYSE.2023.10055143
   
  
• A review of hybrid collaborative filtering algorithms for ELT resources under cognitive diagnosis price  
  by Xianghong Tang 
  Abstract: The study takes English exercises in English teaching resources as the starting point, combines cognitive diagnosis theory to assess students’ knowledge and ability levels. On the basis of integrating the traditional collaborative filtering algorithm, the sorting learning method is introduced, and the combination of the two becomes a hybrid collaborative filtering algorithm. The results show that the accuracy of the proposed hybrid collaborative filtering algorithm under cognitive diagnosis is as high as 98%, with stable performance in accuracy, FI value and recall rate, and all outperform the collaborative filtering algorithm, providing learners with English teaching resources that are more in line with their cognitive ability: the English exercises recommended by the algorithm have better learning effects than those recommended by the collaborative filtering algorithm, effectively providing learners with personalised In practice, the English exercises recommended by the algorithm were more effective than those recommended by the collaborative filtering algorithm.
  Keywords: cognitive diagnosis; sequencing learning methods; hybrid collaborative filtering algorithms; English language teaching.
  DOI: 10.1504/IJCSYSE.2023.10055410
   
  
• Research on e-commerce neural network financial accounting crisis early warning model combined with partial least squares  
  by Xiaoyang Meng 
  Abstract: The study establishes a variable system based on the financial accounting crisis early warning theory, uses partial least squares method to screen variables in order to accurately predict various incentives for the financial crisis in the actual operation of enterprises in the e-commerce industry. According to the findings of the experiment, when the quantity of hidden layer nodes in L-1~3 years is 9, 10 and 11 respectively, the convergence rate of the model can reach the best state; In the prediction of 2020 and 2021, the accuracy rate of L-2 and L-3 is less than 90%, and L-1 has an accuracy rate of more than 90%. In conclusion, the PLS-BP financial crisis early warning model developed and studied can be highly accurate and useful, and it can quickly identify financial crisis signals for businesses in the e-commerce industry and develop efficient measures.
  Keywords: partial least squares; PLS; BP neural network; financial crisis; logistic regression; online retailers; variable indicators.
  DOI: 10.1504/IJCSYSE.2023.10055585
   
  
• A study on the application of teaching differential equations in higher mathematics based on visual network topology algorithm  
  by Li Li 
  Abstract: The effect of teaching evaluation of differential equations in higher mathematics involving network visualisation is extremely advantageous in a multidimensional evaluation system. The study compares the teaching idea of higher mathematical differential equations to a signal flow diagram in a network topology, with mathematical variables as branch nodes for visual structural presentation. A force-guided layout algorithm is introduced to avoid crossover and overlap of nodes. A grey wolf optimisation algorithm incorporating dynamic weights is also used to prioritise the mathematical calculations in conjunction with the features of mathematical differential teaching. The results of the algorithm performance tests showed that the IGWO-visualised layout algorithm had the best optimisation of the functions, with an average optimisation time of 1.6874 s, while the force-guided layout algorithm had an average optimisation time of 12.5986 s.
  Keywords: visual networks; force-guided layout algorithms; IGWO; teaching; topology.
  DOI: 10.1504/IJCSYSE.2022.10055659
   
  
• A study on the impact of personalized recommendation algorithms in webcasting on the development of rural e-commerce entrepreneurship  
  by Jie Li 
  Abstract: The vigorous development of rural e-commerce has brought a great positive effect on the development of rural economy and the improvement of local people’s living material conditions. Among them, online live broadcasting provides a new perspective for the personalised development of e-commerce entrepreneurship. At the same time, based on the advantages of collaborative filtering (CF) algorithm in formulating user scoring criteria, the random forest (RF) algorithm is introduced to realise the research on the correlation of some features, so as to improve the information anti-noise ability and optimise the algorithm performance. And the intelligent recommendation algorithm that combines RF algorithm and improved CF algorithm is applied to rural e-commerce entrepreneurship recommendation. The results show that. The results show that the fusion algorithm combines the advantages of the RF algorithm and the improved CF algorithm, which makes it have better performance in content recommendation.
  Keywords: collaborative filtering; rural e-commerce; random forest; webcasting.
  DOI: 10.1504/IJCSYSE.2022.10055665
   
  
• A comprehensive survey on recommender system techniques  
  by Thenmozhi Ganesan, R. Anandha Jothi, Palanisamy Vellaiyan 
  Abstract: The recommender system (RecSys) is a relatively emergent research area in machine learning that helps users to get personalised products, friends, documents, places and other online services with minimal time. RecSys has been proved as an important solution for information overload problems, by providing more proactive and personalised information services. It performs like a gateway for users to be recommended as to what decision would be right and predicts future post decision. RecSys utilised to support the venture to implement one-to-one marketing strategies in e-commerce. These strategies present enormous advantages namely satisfying the customer's interest increase the possibility of cross-selling and demonstrating the customer loyalty. This paper presents the overview of recommender system approaches, applications and challenges and directly supports the researchers in their understanding of this field. Further, we surveyed collaborative filtering-based RecSys in detailed manner and scrutinised the strengths and limitations to assist the future researchers.
  Keywords: recommender system; RecSys; machine learning; information overload; personalised recommendation; classification and prediction.
  DOI: 10.1504/IJCSYSE.2023.10055735
   
  
• A study of personalised recommendation methods for multimedia ELT online course  
  by Siyi Chen, Xinli Ke, Xiaohong Zhang 
  Abstract: Aiming at the problem that a large number of online courses lead to the reduction of students' efficiency in finding suitable courses, the collaborative filtering recommendation algorithm is improved. The user project scoring matrix is used to calculate the reasonable scoring factors. At the same time, the user project scoring matrix and project characteristics are used to establish a composite feature matrix. Then, combined with demographic information, a mixed user model is established to obtain a neighbourhood set close to the real situation, and finally the best recommendation result is generated. The improved hybrid user model collaborative filtering algorithm (IHUMCF) is used for personalised recommendation. Compare IHUMCF with HUMCF and UBCF. The results showed that IHUMCF recommended the most in the same time; IHUMCF has the highest accuracy rate, recall rate and comprehensive evaluation index, and the lowest average error. It shows that collaborative filtering recommendation algorithm based on improved hybrid user model can improve the accuracy of personalised recommendation, provide better recommendation effect, and analyse students potential learning needs to provide students with a better online learning environment
  Keywords: collaborative filtering algorithm; multimedia English teaching; online courses; personalised recommendation.
  DOI: 10.1504/IJCSYSE.2023.10055787
   
  
• Blockchain Technology: A tool to solve the challenges of education sector in developing countries  
  by Md Aminul Islam, Shabbir Ahmed Shuvo 
  Abstract: Education is getting diversified, challenged, and blended with the overwhelming advancement of modern technology, which can be resolved using blockchain technology (BT). The fourth industrial revolution (4IR) is changing our experiences teaching and learning. Delivering lectures, interacting between learners and educators, evaluating learning outcomes, and verifying educational credentials might be smoother, easier, faster, cheaper, and jollier than before. We have demonstrated that blockchain technology can contribute to the education provider tackling all those existing problems to create a comfortable learning environment for all, irrespective of their economic backgrounds and geographic location. How BT can contribute to improving education is one of our priorities of research in this study. Firstly, this study reviews recent inventions in BT. Secondly, we have gone through the connection between BT and education. Additionally, it discusses strategies around the world. Few models are arranged to enable the reader's mind to inventions in the realm of educationists.
  Keywords: blockchain; 4IR; educators; learning outcome; blockchain in education; education; sustainable education; modern education; challenges of education.
  DOI: 10.1504/IJCSYSE.2023.10055995
   
  
• Research on credit risk assessment of e-commerce enterprises based on improved multi-objective clustering algorithm  
  by Danyan Zhong 
  Abstract: With the increasing share of e-commerce business, many companies are also facing credit risk issues of varying degrees. Aiming at the problem of credit risk assessment, this study proposes an improved multi-objective clustering algorithm to assess corporate credit risk. By comparing the conventional FMC algorithm and K-means algorithm, the performance of the proposed improved MOEC algorithm is analysed. It can be seen from the PR curves of the three algorithms that the AP values of the three algorithms are 0.9324, 0.9455, and 0.9972, respectively. In contrast, the improved MOEC algorithm has higher accuracy and stability. Tested on the UCI dataset, it was found that in dataset 3, the RI value of the improved MOEC algorithm was 0.97; in dataset 5, the NMI value of the improved MOEC algorithm was 0.96.
  Keywords: e-commerce; corporate credit; clustering algorithm; multi-objective; risk assessment; finance.
  DOI: 10.1504/IJCSYSE.2023.10056099
   
  
• Machine Learning in Financial Risk Forecasting and Management for Trading Firms  
  by Zhiqing Zhou 
  Abstract: To improve the financial risk prediction ability of e-commerce enterprises, this study combines BP neural network and PLS method to construct a financial crisis warning model for e-commerce enterprises, namely the BP-PLS model. The experiment first analyses and selects financial crisis warning indicators for e-commerce enterprises, and then extracts components using PLS method. Then, the results of component extraction are used as input vectors for the BP neural network. Finally, the experiment used the BP-PLS model to construct financial crisis warning models for e-commerce enterprises in T-1, T-2, and T-3 years, respectively. The experimental results show that the accuracy of both T-1 and T-2 models is above 90%. The accuracy of the T-3 model exceeds 85%. Therefore, the established model can meet the needs of financial crisis warning. In addition, the model has excellent performance and its training error convergence effect is superior to other models.
  Keywords: machine learning; e-commerce enterprise; financial risk; forecast; BPNN.
  DOI: 10.1504/IJCSYSE.2023.10056235
   
  
• The effects of integrated feedback based on AWE on English writing of Chinese EFL learners  
  by Mei Liu, Changzhong Shao 
  Abstract: Through a two-semester experiment on the English writing of 64 Chinese EFL learners, this study examines the effects of three types web-based feedback [automatic feedback (AF), AF with teacher feedback (TF), AF with peer feedback (PF)] based on Pigai website. The results show that all the three modes of feedback can promote the writing of the English as foreign language (EFL) learners with different English proficiency. The results also reveal that AF with TF is more effective in helping the learners with higher English proficiency improve their writing capacity, while for the learners with lower English proficiency, AF with TF works best in enhancing their writing capacity.
  Keywords: integrated feedback; automated writing evaluation; AWE; English writing; Pigai website; English as foreign language; EFL.
  DOI: 10.1504/IJCSYSE.2024.10056238
   
  
• Analysis of word vector combined with group intelligence perception based on STEM concept for ELT word recommendation strategy  
  by Ke Qin, Yang Zhou, Daniel Kvitrud 
  Abstract: With the rapid development of China's education concept, the mode of English learning has produced great changes, and the learning methods of ELT words have gradually developed towards intelligence. For the problem of ELT word recommendation strategy, an optimised CWSAR algorithm model based on STEM concept, combining collaborative filtering algorithm, word vector semantic perception, context perception and crowdsensing is proposed. The optimal parameter experiments and comparison experiments are conducted for this algorithm model. The experimental results show that CWSAR algorithm is better than the two CF algorithms in the experiment of ordinary data sets. The CWSAR algorithm can better complete the work of English teaching word recommendation and provide effective help for English teaching.
  Keywords: word recommendation; collaborative filtering; word vector; group wisdom perception.
  DOI: 10.1504/IJCSYSE.2024.10056267
   
  
• The construction of college students' job recommendation model based on improved k-means-CF  
  by Ping Ouyang 
  Abstract: Based on the Internet of Things (IoT) technology, building a digital management platform for employment and entrepreneurship service system, recommending suitable corporate positions for students and promoting students’ employment and entrepreneurship has become an important issue for each university. At present, the recommendation accuracy and recommendation efficiency of most digital management platforms of employment and entrepreneurship service system are not ideal and not very practical. To this end, the research is based on the idea of data mining, combining collaborative filtering algorithm (CF), k-means algorithm and dichotomous k-means algorithm to build a personalized recommendation model for graduate jobs, and improve and optimize the career recommendation of the digital management platform of employment and entrepreneurship service system based on this model. The experimental results show that the accuracy rate of model 4 reaches 99.78%, which is significantly higher than the other three models. Therefore, the personalised recommendation model constructed by the study can efficiently and accurately provide students with employment and entrepreneurship information, thus promoting students employment and entrepreneurship and providing some relief to the huge employment pressure in the current society.
  Keywords: Internet of things; employment entrepreneurship; k-means algorithm; collaborative filtering algorithm; dichotomous k-means algorithm; data mining.
  DOI: 10.1504/IJCSYSE.2023.10056610