International Journal of Web Engineering and Technology (13 papers in press)

Regular Issues

• Service Recommendation Method based on Text View and Interaction View  
  by Shuaijia Lin, Ting Yu, Yaqi Wang, Jie Xu, Fangying Cheng, Tian Liang 
  Abstract: With the increasing prosperity of web service-sharing platforms, more and more software developers are reusing web services when developing applications. Existing web service recommendation systems often face two challenges. Firstly, developers discover services by inputting requirements, but the user's input is arbitrary and it cannot fully reflect the user's intention. Secondly, the application-service interaction records are too sparse, making it particularly difficult to find services that meet the requirements. To address the above challenges, in this paper, we propose a service recommendation method based on text and interaction views (SRTI). Firstly, SRTI employs graph neural network to deeply mine the features of applications and services. Secondly, SRT uses transformer and fully connected neural networks to deeply mine the matching degree between candidate services and requirements. Finally, we integrate the above two to obtain the final service list. Extensive experiments on real-world datasets have shown that SRTI outperforms several state-of-the-art methods.
  Keywords: service recommendation; text view; interaction view; application; recommendation algorithm.
  DOI: 10.1504/IJWET.2024.10064249
   
  
• Authentication of Musical Speech Devices Based on RF Fingerprint Recognition  
  by Zitian Liao, Xiaoqun Liao 
  Abstract: Music-voice devices are becoming more and more abundant and diverse with the development of technology; their security and privacy issues have also attracted widespread attention Therefore, the research aims to explore the authentication method of music voice devices based on RF fingerprinting, and to propose an efficient and accurate authentication method combining the time-domain features and frequency-domain features of RF signals through in-depth analysis of RF signal features The research results indicated that the feature fusion RF fingerprint identification technology is significantly better than the single algorithm, deep learning and neural network methods in terms of accuracy, precision rate, operation efficiency and anti-interference ability.The research has practical applications in enhancing the security of music voice devices, protecting user privacy, and improving user experience. It is significant in promoting the development of RF fingerprint identification technology.
  Keywords: Radio frequency fingerprint identification; Musical voice devices; Authentication; Feature fusion.
  DOI: 10.1504/IJWET.2025.10068323
   
  
• Fusion of Visual Attention Model and SVM for Sentiment Analysis of Planar Images  
  by Yongsong Liu 
  Abstract: The sentiment analysis of images is actually a classification problem. However, a common classifier is unable to handle the emotion classification of images. This study proposes a planar image sentiment analysis recognition method that simulates the human eye recognition process to enhance the accuracy of sentiment analysis in planar images. This method utilises an improved block wise adaptive weighted colour histogram and a cognitive feature extraction model that integrates visual attention models to extract features from images. Finally, this method combines several extracted features of the image and uses support vector machines for sentiment recognition and classification. The results showed that when the words in the dictionary were 90, the algorithm performed the best, with the highest accuracy of 69%. In emotion recognition classification, the more image features were fused, the better the emotion classification recognition effect of the image. When integrating four features, the highest classification accuracy of the algorithm was 72.9%. The research proposes a planar image sentiment analysis method that integrates visual attention models and support vector machines, which can effectively simulate the changes in the focus area features of the human eye when observing images. This can achieve accurate recognition of emotions expressed in images.
  Keywords: Visual attention; SVM; Flat image; Emotional analysis; Feature extraction.
  DOI: 10.1504/IJWET.2025.10068655
   
  
• Digital Education Mining Technology Based on Composite Collaborative Filtering and Eclat Algorithm  
  by Jingya Wang, Qi Han, Kunkun Ma, Li Xu 
  Abstract: The field of digital education is rapidly growing, demanding effective resource utilisation. Traditional collaborative filtering (CF) algorithms face challenges with large, complex datasets. This study addresses these limitations by integrating CF with association rule mining, using a novel IBCF-UBCF composite CF algorithm and Eclat technology. Data was collected from multiple sources and fused for enhanced educational mining. Results show Eclat outperforms apriori, reducing CPU usage by 55% and physical memory usage by 51.9%, while the composite filtering algorithm achieved over 99% accuracy. The Eclat-IBCF-UBCF algorithm offers robust support for digital education, advancing educational data mining and personalised recommendations. It is recommended for implementation in digital education systems due to its efficiency and accuracy. Further research should focus on enhancing and integrating this algorithm with other educational technologies.
  Keywords: Eclat; IBCF algorithm; UBCF algorithm; Composite collaborative filtering; Digital education.
  DOI: 10.1504/IJWET.2025.10068762
   
  
• Clustrosearch: A Novel, Intent-Aware, and Anomalous Reducing Meta-Search Engine Optimisation Algorithm based on User Scoring System and Clustering  
  by Parsa Parsafar 
  Abstract: This paper introduces Clustrosearch, a novel meta-search engine optimization algorithm integrating a machine learning-based user scoring system to enhance search result accuracy and efficiency. In response to the shortcomings of traditional search engines in delivering tailored content, Clustrosearch innovatively addresses these challenges by prioritizing user-centric information retrieval. Unlike existing approaches, it focuses on optimizing result relevance without explicit intent awareness, making it versatile across various search scenarios. Clustrosearch incorporates advanced anomaly reduction techniques to minimize the impact of outlier results, thereby enhancing the overall quality of search outcomes. This approach is evaluated through comprehensive benchmarking against established meta-search algorithms, demonstrating its capability to significantly reduce irrelevant results and improve retrieval precision. This research underscores its scalability and effectiveness in enhancing user satisfaction through advanced information filtering techniques.
  Keywords: Search Engine Optimization (SEO); Meta-Search Engine Optimization; Optimization; Linear Search; Rank-Biased Overlap (RBO).
  DOI: 10.1504/IJWET.2025.10068922
   
  
• A Sanda Action Recognition Using CNN-LSTM Network Model  
  by Jingying Ouyang, Jisheng Zhang, Yuxin Zhao, Shenghai Chen 
  Abstract: This study presents an action recognition algorithm based on convolutional long short-term memory (CNN-LSTM) to enhance the movement analysis precision. The model takes the joint position as the recognition node, and combines with the cylindrical coordinate system with random sliding window to effectively capture the angle and position information of the action frame. The algorithm groups joints, selects root nodes, and extracts local features through a multi-stream network, with classification completed in the pooling and fully connected layers. The proposed model achieves an average accuracy of 98.89%, with a recognition time of 0.61s and a minimal deviation of 0.035 in Sanda movements, demonstrating superior performance in action recognition.
  Keywords: Action recognition; Convolution; Long short-term memory network; Joint; Sanda.
  DOI: 10.1504/IJWET.2025.10068998
   
  
• Smart Campus Integration System based on SLAM and Digital Twin Drive Technology  
  by Weihua Feng 
  Abstract: Building a smart campus integration platform can help colleges and universities solve the difficult and painful problems in the early stage of informatization construction, such as inconvenient sharing of educational resources, poor experience of service system, insufficient utilisation rate of equipment, and risks of information security In order to improve the performance effect of smart campus system in university operation and teaching management, this paper establishes a standard information model of smart campus data, and adopts structured data to fuse data to improve the efficiency and accuracy of the system Moreover, through the integration of big data and digital twin technology, this paper solves the problems of information island, real-time, intelligent service and poor experience in university information system by combining twin data for virtual and real-time integration and real-time interaction In addition, this paper verifies the effect of the smart campus integration system through experimental verification methods The experimental results show that the comprehensive performance of this system is strong, so it can be seen that the system proposed in this paper is reliable and effective in the smart campus integration system, and the smart campus system can be continuously improved through data mining and digital twinning technology in the follow-up research.
  Keywords: data mining; digital twinning; drive; smart campus; integration system.
  DOI: 10.1504/IJWET.2025.10069445
   
  
• Combined Jelly-Snake Optimisation with Deep Learning Architecture for Task Offloading and Resource Allocation in Edge Computing  
  by RAJA A, Prathibhavani P. M, Venugopal K. R 
  Abstract: Edge computing allows devices to transfer their computational tasks to nearby edge servers. However, effectively managing offloading decisions and optimising resource remains challenge. To tackle this issue, this paper proposes an innovative method for task offloading and resource allocation in edge computing. In this, CNN-based task offloading method is utilised, incorporates WBAN, ES, and a medical center integrated into edge computing for offloading structure. The GUN handles task-related data scaling, which serves as input for CNN. The CNN produces result ranging from zero to one, where "0"-local task execution and "1"- task offloading to ES. Then, hybrid algorithm combines JSO and SOA is proposed to effectively manage workload demands. The SUJO method optimises resource allocation by considering factors-makespan, task priority, execution time, and energy consumption. The comparative analysis demonstrates SUJO's superiority, achieving execution time of less than 50 seconds and proves effective for optimising task offloading and resource allocation.
  Keywords: WBAN; GUN; Edge computing; Resource allocation; Task offloading; CNN; SUJOA.
  DOI: 10.1504/IJWET.2025.10069446
   
  
• Optimisation Algorithm for Distributed Economic Dispatching Problems in Smart Grid  
  by Yan Li, Zhibang Ruan 
  Abstract: The development of smart grids faces many problems. The urgent problem is the dispatch of smart grids. To address this, a smart grid economic dispatch model based on a second-order consistency algorithm is proposed. The model transforms dispatch into an unconstrained optimisation problem using convex optimisation theory and integrates a multi-agent consistency algorithm with the internal penalty function method. According to the experimental results, when the training set was 900, the energy saving efficiency of the traditional dispatch model, the dispatch model based on the first-order consistency algorithm, and the second-order consistency algorithm during peak electricity consumption periods were 0.38, 0.55, and 0.64, respectively. During the low power consumption period, when the training set size was 900, the energy saving efficiency of the three models was 0.32, 0.41, and 0.49. The proposed method can effectively solve the power grid dispatch issues, providing a certain reference for relevant research.
  Keywords: Smart grid; Economic dispatch; Distributed second-order consistency algorithm; Internal penalty function method.
  DOI: 10.1504/IJWET.2025.10069681
   
  
• Industrial Design Pattern Optimisation Research based on 3D Deep Learning algorithm  
  by Yao Wang 
  Abstract: The appearance design of industrial products is mainly done manually, which is time-consuming and inconsistent. To improve efficiency and quality, an intelligent design model based on Generative Adversarial Networks (GAN) was developed. The model converts 3D input data into a scene diagram to represent component relationships and hierarchical structures, making it suitable for complex product layouts. By incorporating the variational autoencoder (VAE) principle, the richness of design information is enhanced. GraphVAE is utilised to learn graph features and generate new structures that meet industrial design needs. Experimental results show that with 1000 test samples, the average information entropy of the proposed model, GoogLeNet, and Fast R-CNN is 35.79, 26.17, and 31.25, respectively, indicating high information richness and similarity to input data. This method optimises the design process, enhances output quality, and reduces computational costs, though its efficiency is lower, making it suitable for less time-sensitive applications.
  Keywords: Keywords: Industrial products; Appearance design; GAN; Artificial intelligence.
  DOI: 10.1504/IJWET.2025.10069713
   
  
• Sentiment Analysis using Deep Learning Algorithms Based on Chat Records and Product Reviews  
  by Haili Lu, Lin He 
  Abstract: With globalisation and the popularity of the internet, consumer evaluation and product feedback are no longer limited to traditional channels, but expressed through online chat and product reviews. These comments contain a wealth of emotional information and viewpoints, which are of great value to the enterprise. In response to the difficulty of traditional sentiment analysis models in handling complex emotional expressions and semantic information, a method combining support vector machines with bidirectional long short-term memory networks is proposed. The experimental results show that the average classification error of this model is less than 2.4% on the LAMAZON and Yelp datasets, which is superior to other schemes. In contrast, the fitting degree of this model is 99.7%, ranking the highest among all algorithms, and the accuracy of emotion classification exceeds 90%. Therefore, the model combining SVM and BiLSTM performs well in sentiment analysis tasks with high accuracy, providing valuable decision support for enterprises.
  Keywords: Support vector machine; Long short-term memory network algorithm; Sentimental analysis techniques; Chat records; Product reviews.
  DOI: 10.1504/IJWET.2025.10069715
   
  
• Assessment of Efficient Machine Learning Algorithms for Enhancing Road Safety and Predicting Accident Severity  
  by Akshi Bharadwaj, Sudesh Kumar, Pawan Singh 
  Abstract: In response to the pressing need for road safety enhancement, this study explores the implementation of Machine Learning (ML) methods to forecast the severity of accidents. The research aims to uncover the intricate factors underlying accidents and provide actionable insights for proactive measures. Utilising algorithms including Random Forest, Support Vector Classification (SVC), XGBoost, Balanced Bagging, and Voting Classifier, the study evaluates performance using standard metrics such as F1 score, recall, precision, and accuracy. Notably, the Soft Voting Classifier, comprising XGBoost, Balanced Bagging, and Gradient Boosting, emerges as the leading model with an accuracy rate of 85.7%, demonstrating its efficacy in accident severity prediction.
  Keywords: Accident predictions; Machine learning; ITS; XGBoost; Random Forest.
  DOI: 10.1504/IJWET.2025.10069725
   
  
• Enhancing User Intimacy: a Parallel Mining Algorithm Utilising UISFW  
  by Ruiping Kong, Ruimin Pan 
  Abstract: Abstract: User intimacy is vital for gauging the emotional bonds and interactions between users and brands, products, or services, crucial for enhancing user experience, brand value, marketing efficacy, and fostering innovative enterprise growth. Existing algorithms for user intimacy mining struggle with accuracy and potential relationship identification, limiting practical usability. Proposing a refined method, user intimacy calculation integrates genetic algorithms, enhancing relationship analysis. Through this approach, a parallel mining model employing Mad Reduce demonstrates promising results. Achieving P@n values of 90.28% and 90.89%, and NDCG values of 82.88% and 82.73% across datasets, the model effectively computes user intimacy and uncovers valuable insights for user relationship identification, benefiting product design and information recommendations.
  Keywords: Keywords: User intimacy; Feature assignment; Genetic algorithm; Parallel mining; MapReduce.
  DOI: 10.1504/IJWET.2026.10070867