International Journal of Web Based Communities (24 papers in press)

Special Issue on: Automatic Disinformation Detection on Social Media Platforms

• Data mining method of social media hot topics based on time series clustering  
  by Wei Wang 
  Abstract: In order to overcome the problems of large analysis error and low mining accuracy of traditional hot topic data mining methods, this paper proposes a new social media hot topic data mining method based on time series clustering. Firstly, the topic feedback forms of reading, comment, forwarding and praise are taken as the research objects. Secondly, the contribution value of various data in the topic heat is calculated to obtain the topic heat results. Finally, take the topic value as the goal and follow-up reports as the index to realise data mining. The test results show that the design method can accurately analyse the data of reading, comments, forwarding and likes. The analysis results of the number of follow-up reports have a high degree of fit with the actual results, and have a high mining accuracy, which is close to 100%.
  Keywords: time series clustering; social media hot topics; data mining; topic heat; topic value.
  DOI: 10.1504/IJWBC.2023.10049291
   
  
• Prediction method of e-commerce consumers' purchase behavior based on social network data mining  
  by Ming Yang 
  Abstract: In order to effectively improve the prediction accuracy of e-commerce consumers’ purchase behaviour and shorten the prediction time of e-commerce consumers’ purchase behaviour, a prediction method of e-commerce consumers’ purchase behaviour based on social network data mining is proposed. Firstly, according to the statistical characteristics of e-commerce consumers’ purchase behaviour, data mining method is used to extract the characteristics of e-commerce consumers’ purchase behaviour. Secondly, the social network analysis method is used to analyse the purchase behaviour characteristics of e-commerce consumers and the social network model. Finally, build the prediction model of e-commerce consumers’ purchase behaviour to realise the prediction of e-commerce consumers’ purchase behaviour. The experimental results show that the proposed method has a good effect on the prediction of e-commerce consumers’ purchase behaviour, and can effectively improve the prediction accuracy of e-commerce consumers’ purchase behaviour. The prediction deviation rate is only 1.8%.
  Keywords: social network analysis method; data mining methods; e-commerce consumers; purchase behaviour; prediction model.
  DOI: 10.1504/IJWBC.2023.10049292
   
  
• Cross modal retrieval of large-scale images in social media based on spatial distribution entropy  
  by Jie Ding, Guotao Zhao, Fang Xu 
  Abstract: In order to improve the cross-modal retrieval accuracy of large-scale social media images, a cross-modal retrieval method for large-scale social media images based on spatial distribution entropy is proposed. First, extract the information features of the colour and texture of the image, then use the image cross-modal retrieval method based on the spatial distribution entropy to calculate the spatial distribution entropy of the colour information and texture information features in the image, and finally use the Euclidean distance to judge the space between social media images. The matching degree of the distribution entropy, according to the matching degree to judge whether the image cross-modal retrieval is successful or not. The experimental results verify that the proposed method can implement comprehensive retrieval according to the specific characteristics of the retrieved images, and the matching degree of image retrieval is greater than 95%, and the retrieval accuracy is high.
  Keywords: spatial distribution entropy; social media; large-scale; cross-modal; image; retrieval.
  DOI: 10.1504/IJWBC.2023.10049668
   
  
• Data mining method of mobile e-commerce consumer purchase behavior  
  by Shuang Li 
  Abstract: In order to improve the accuracy and efficiency of data mining of consumer purchase behaviour in mobile e-commerce, a data mining method of consumer purchase behaviour in mobile e-commerce is proposed. Firstly, through the calculation of support in principal component analysis, the data characteristics of mobile e-commerce consumers’ purchase behaviour are extracted. Then, the original residual of purchase behaviour data is calculated through iterative test, and the original feature data is corrected to complete the preprocessing. Finally, the Boolean rules in association rules are used to determine the association degree between purchase behaviour data, and the minimum threshold of purchase behaviour data is calculated. By establishing the correlation function of mobile e-commerce consumer purchase behaviour data, mining the characteristic information of mobile e-commerce consumer purchase behaviour data, and completing the purchase behaviour data mining. The results show that the highest accuracy of data mining is 98.1%.
  Keywords: principal component analysis; iterative test method; Boolean rule; mobile e-commerce; consumer buying behaviour.
  DOI: 10.1504/IJWBC.2023.10050219
   
  
• An Encryption of Social Network User Browsing Trajectory Data Based on Adversarial Neural Network  
  by Xinliang Wang 
  Abstract: In order to solve the problems of high information loss rate, poor encryption effect and long encryption time existing in traditional social network user browsing trajectory data encryption methods, this paper proposes an encryption method of social network user browsing trajectory data based on adversarial neural network. Mutual information is used to extract browsing characteristics of social network users and calculate browsing path similarity of social network users, so as to determine the clustering centre of browsing trajectory data and realise browsing trajectory data mining. Combining with adversarial neural network, the symmetric encryption and decoding model is designed, and the user browsing feature data is input into the model to realise the user browsing feature data encryption. Experimental results show that the information loss rate of the proposed method is always lower than 5%, the encryption effect is good, and the average encryption time is 53 ms.
  Keywords: adversarial neural network; social network; user browsing trajectory; data encryption; data mining; symmetric encryption.
  DOI: 10.1504/IJWBC.2023.10050220
   
  

Special Issue on: Decentralisation and New Technologies for Social Media

• A personalized recommendation method of online educational resources on social media platform  
  by Ziqian Xu 
  Abstract: Aiming at the problems of low recommendation accuracy and low user preference in traditional methods, a personalised recommendation method of online educational resources on social media platform is proposed. Firstly, the crawler technology is used to obtain the online educational resources data, and the resource data features are extracted; Then, the similarity of data features is calculated through cosine similarity, and the feature data with high similarity is fused to complete the feature preprocessing of educational resources data; Finally, the user’s demand for resource data and preference degree are determined through the user interest model, so as to construct the online educational resources personalised recommendation model, and take the educational resources data and user preference degree as the input data to complete the educational resources personalised recommendation. The experimental results show that the proposed method has high accuracy and user preference.
  Keywords: social media platform; online educational resources; personalised recommendation; reptile technology; interest model.
  DOI: 10.1504/IJWBC.2023.10047482
   
  
• Study on novel Cross-chain Mechanism in Internet Healthcare Environment  
  by Meiyan Wei, Haiying Wen, Danlei Du, Wei Ou 
  Abstract: The scalability and interoperability of traditional blockchain can hardly meet the value delivery of multiple systems in internet healthcare. Therefore, to enable the interaction of values between heterogeneous blockchains, we propose a cross-chain mechanism based on a generic global relay chain strategy and a scalable heterogeneous multi-chain system. Also, an unused transaction output (UTXO) verification mechanism based on RSA accumulator is proposed to address the shortcomings of using simplified payment verification (SPV) to verify transactions. In this paper, five experimental tests were performed: pressure, stability, response time, throughput, and latency. Experiments show that its computational complexity for cross-chain transaction validation remains constant as the number of transactions increases. Thus, an efficient cross-chain value transfer is achieved.
  Keywords: internet healthcare; data silos; cross-chain; relay chain; consensus.
  DOI: 10.1504/IJWBC.2023.10048309
   
  
• Personalized advertising Push method based on Semantic similarity and data mining  
  by Shibiao Mu, Shuaijing Yu 
  Abstract: This paper designed a personalised advertising push method based on semantic similarity and data mining. Firstly, in order to improve the matching degree of advertising keywords, the similarity theory is used to classify advertising categories. According to the classification results, search engine technology is used to match user preferences and advertising keywords to increase the matching degree between advertising content and users. Finally, on the basis of determining the target advertising project, the ads with high semantic similarity are pushed to users as the results. The results show that the matching degree of advertising keywords in this method is between 85% and 95%, the highest accuracy of advertising classification can reach 94%, and the user satisfaction is the highest, indicating that this method has greatly improved the effect of advertising push.
  Keywords: semantic similarity; data mining; advertising push; search engine technology; key word; association rules.
  DOI: 10.1504/IJWBC.2023.10048310
   
  
• An accurate identification method of abnormal users in social network based on multivariate characteristics  
  by Jian Xie 
  Abstract: In this paper, an accurate identification method of abnormal users in social networks based on multiple features was proposed. Firstly, the API interfaces provided by social networks are used to capture social network user data, so as to extract multiple features such as account features, text features and behaviour characteristics of users. Then, attribute reduction method is used to remove redundant features and obtain accurate user attribute feature set. Finally, based on the user attribute feature set, XGBoost model was used to construct the objective function of accurate identification of abnormal users in social networks, and the accurate identification results of abnormal users in social networks were obtained. Experimental results show that the feature extraction accuracy of abnormal users in social networks by the proposed method is more than 95%, the identification error rate varies between
  Keywords: multivariate characteristics; social networking; abnormal users; accurate identification; attribute reduction; XGBoost model.
  DOI: 10.1504/IJWBC.2023.10048311
   
  
• Design of social media information extraction system based on deep learning  
  by Huimin Wang, Yaping Gao 
  Abstract: Aiming at the problems of low accuracy and long time in traditional systems, a social media information extraction system based on deep learning is designed. Firstly, the overall framework of the system is designed, including text extraction module, keyword extraction module and emotion analysis module. Then, the social media information is preprocessed, the emotional resource establishment and information extraction rules are constructed according to the preprocessing results, and the convolution neural network is used to construct the social media information extraction model. Finally, according to the correlation between text entries and categories, the global MI values of entries and all categories are calculated, the calculation results are input into the constructed convolution neural network model, and the social media information extraction results are output. The simulation results show that the extraction accuracy of the designed system is high and the extraction time is within 15 s.
  Keywords: social media; information extraction; emotional resources; convolutional neural network; text entry.
  DOI: 10.1504/IJWBC.2023.10048312
   
  
• A comprehensive retrieval method of social media information based on Fuzzy Mathematics  
  by Xinyan Yu 
  Abstract: In order to solve the problems of low retrieval accuracy and long retrieval time of traditional information comprehensive retrieval methods, a social media information comprehensive retrieval method based on fuzzy mathematics is proposed in this paper. Firstly, the social media information data is collected by hash function; Then, the cross filter is used to remove similar social media data; Finally, according to the fuzzy triangle number in the fuzzy mathematics method, the keyword weight in social media information is calculated, and the comprehensive retrieval of social media information is completed by determining the attribute of social media information and reducing the word frequency range of social media information. The experimental results show that the minimum detection error of this method is about 0.15%, and the shortest information retrieval time is about 1.0 s. The retrieval effect of the proposed method is good.
  Keywords: fuzzy mathematics method; social media information; comprehensive search; hash function; high level semantics; fuzzy mathematics.
  DOI: 10.1504/IJWBC.2023.10048313
   
  
• Dynamic monitoring of network public opinion diffusion of major public crisis based on deviation degree  
  by Jianxin Qiu 
  Abstract: In order to solve the problems existing in traditional methods such as low accuracy and long monitoring time, a dynamic monitoring method of network public opinion diffusion of major public crisis based on deviation degree is proposed. Firstly, the crawler is used to collect network public opinion data, and the collected data is preprocessed. Secondly, the commonly used query keywords in search engines are classified into statistics, and the diffusion anomalies of network public opinions on major public crisis are monitored with the deviation degree. Finally, according to the monitoring results, a report on the network public opinion of major public crisis is generated to realise the dynamic monitoring of the spread of network public opinion of major public crisis. Experimental results show that the designed dynamic monitoring method has the characteristics of high accuracy, low deviation, short monitoring time and high reliability, and the practical application effect is good.
  Keywords: deviation; major public crisis; network public opinion diffusion; dynamic monitoring.
  DOI: 10.1504/IJWBC.2023.10048314
   
  

Special Issue on: Information Technology and Consumer Behaviour Challenges and Opportunities

• Predictive model of consumer online purchase behavior based on data mining  
  by HaoJie Zi 
  Abstract: The existing consumer online purchase behaviour prediction model does not reduce the noise of purchase behaviour data, which leads to poor prediction effect. Therefore, a consumer online purchase behaviour prediction model based on data mining is proposed in this paper. The behaviour data are divided into different datasets by K-means clustering; The neighbourhood rule is used to update the centre of clustering sample data and collect behaviour characteristic data. Empirical mode decomposition (EMD) method is used to obtain the instantaneous frequency of purchasing behaviour and denoise the characteristic data of consumers’ online purchasing behaviour; With the data mining method, the system of consumer online purchasing behaviour characteristics is established, and a consumer online purchasing behaviour prediction model is built according to the behaviour characteristics fusion selection means to realise behaviour prediction. The results show that the accuracy of this model to predict consumers’ online purchasing behaviour is more than 93%.
  Keywords: data mining; consumer behaviour; purchasing behaviour; behaviour prediction; online shopping.
  DOI: 10.1504/IJWBC.2023.10048315
   
  
• A Rough Set Based Consumer Buying Behavior Prediction Method in Online Marketing System  
  by Dian Jia 
  Abstract: Aiming at the problems of large prediction deviation and low acquisition accuracy of consumer purchase behaviour in traditional online marketing systems, a rough set-based consumer purchase behaviour prediction method in online marketing system is proposed. By improving the accuracy and recall rate of online consumer buying behaviour prediction methods, the deviation of prediction results is reduced. The data of consumer purchase behaviour in the region related to rough set are reduced to improve the accuracy and recall rate, and the forecast bias is reduced by removing redundant features in the e-marketing system. With the rough set theory, the dimension of consumer behaviour vector is reduced, and a predictive model framework is built. The simulation results show that the accuracy and recall rate of this proposed method are higher than 95%, and the minimum deviation of the prediction result is only 8.12%, which proves that the prediction result is more reliable.
  Keywords: rough set theory; online marketing system; consumers; buying behaviour prediction.
  DOI: 10.1504/IJWBC.2023.10048316
   
  
• Research on the Fuzzy Comprehensive Evaluation of Consumer Satisfaction with Mobile E-Commerce Platforms  
  by Xulan He 
  Abstract: Because the traditional method has the problems of a short evaluation time, accurate information retrieval and accurate evaluation, a new fuzzy comprehensive evaluation method for consumer satisfaction with mobile e-commerce platforms is proposed. First, the relevant content of mobile e-commerce platforms is analysed based on the content and functional characteristics of the design rules of fuzzy evaluation. Then, customer satisfaction evaluation principles are analysed, and the corresponding secondary evaluation index is established based on the weight of each index to complete a specific consumer satisfaction fuzzy comprehensive evaluation. The experimental results show that the time of the evaluation result generation process of this method is between 27 s and 33 s, the retrieval accuracy of consumer-related information is between 0.931 and 0.952, and the evaluation accuracy is between 93.0% and 95.6%. The above results effectively prove the effectiveness of this method.
  Keywords: mobile e-commerce platform; consumer satisfaction; fuzzy rules; satisfaction evaluation; index weight.
  DOI: 10.1504/IJWBC.2023.10048317
   
  
• Cluster Analysis of Perceptual Demands of Users' Internet Consumption Behaviors Based on Improved RFM Model  
  by Yuxi ZHang  
  Abstract: In order to overcome the problems of traditional clustering analysis methods, such as low accuracy, long consuming time and less demand types, a clustering analysis method based on improved RFM model is proposed in this paper. The intelligent internet of things platform is used to collect the data of users’ online consumption behaviour, and the frequent patterns of the data collection results are mined according to the big data fusion method; The improved RFM model is used to obtain three parameters of users’ latest consumption, user consumption frequency and consumption amount, so as to realise the clustering analysis of users’ perceptual demand of online consumption behaviour. The experimental results show that with high clustering analysis accuracy and ability of consuming clustering analysis time by always less than 9.0 s, this proposed method can effectively cluster more types of user needs, suggesting that the clustering analysis effect of this method is relatively ideal.
  Keywords: improved RFM model; online consumption behaviour; perceptual demand clustering; smart IoT platform; frequent patterns.
  DOI: 10.1504/IJWBC.2023.10048318
   
  
• Cluster analysis-based big data mining method of E-commerce consumer behavior  
  by Lei Xue  
  Abstract: In order to overcome the problems of low precision and long time of data mining in traditional big data mining methods of consumer behaviour, a clustering analysis method for big data mining of e-commerce consumer behaviour is proposed. In this paper, the K-means algorithm is used to calculate the similarity of behaviour clustering nodes of fee payers, determine the clustering process of consumer behaviour data and determine the mining weight of behaviour data. According to the FCM clustering algorithm, the target function for data mining of e-commerce consumer behaviour is constructed. According to Lagrange multiplication, the membership degree of consumer behaviour data is obtained, and the big data mining of consumer behaviour in e-commerce is realised. The experimental results show that with the method proposed in this paper, when the number of consumers is 500, the time for big data mining of consumer behaviour is 15.6s and the accuracy of behaviour big data mining is 95.34%.
  Keywords: data mining; K-means clustering analysis; FCM clustering algorithm; Lagrange multiplication; cluster node.
  DOI: 10.1504/IJWBC.2023.10048319
   
  
• A deep mining method for consumer behavior data of e-commerce users based on clustering and deep learning  
  by Jing Li 
  Abstract: The data mining accuracy of e-commerce users’ consumption behaviour is low and the data clustering effect is poor, so a deep mining method of e-commerce users’ consumption behaviour databased on clustering and deep learning is proposed. The consumption behaviour data are divided into simple type, deterministic type, habitual row type and preference type through the user’s web browsing log, and the features of the consumption behaviour data are extracted. The centroid and class spacing of behaviour characteristic data are obtained according to the actual distance between the behaviour characteristic data points. The behaviour data deep mining model is built based on the small wave neural network and the deep learning algorithm, and the optimal solution of the model is thus obtained by the gradient descent method, so as to realise the deep mining of the consumption behaviour data. The results show that the accuracy of the proposed method is up to 97%.
  Keywords: data clustering; deep learning; dimension kernel function; centroid.
  DOI: 10.1504/IJWBC.2023.10048320
   
  

Special Issue on: Social Media and Consumer Behaviour Challenges and Opportunities

• Consumer behavior data mining of social e-commerce platform based on improved spectral clustering algorithm  
  by Ru Zhang 
  Abstract: In order to overcome the problems of low recall rate, precision rate and long mining time of traditional methods, a consumer behaviour data mining of social e-commerce platform based on improved spectral clustering algorithm is proposed. Firstly, the collaborative filtering algorithm is used to predict the degree of user preference, and data crawler is designed according to the LDA theme model to crawl the consumer behaviour data of social e-commerce platform, the data is cleaned and processed. Then, the initial clustering centre optimisation algorithm is used to improve the spectral clustering algorithm, and the improved spectral clustering algorithm is used to cluster the data cleaning results to realise consumer behaviour data mining. Finally, the simulation experiment proves that the recall rate and precision rate of the proposed method are both high, and the data mining time is always less than 0.55 s.
  Keywords: improved spectral clustering algorithm; social e-commerce platform; consumer behaviour; data mining; LDA topic model; data crawler.
  DOI: 10.1504/IJWBC.2023.10048828
   
  
• Online consumer behavior anomaly recognition method based on limit learning machine  
  by Zheng Xie, Lianguang Mo 
  Abstract: Aiming at the large identification error and long identification time in online consumer behaviour anomaly identification, an online consumer behaviour anomaly identification method based on limit learning machine is designed. The key factors affecting the characteristics of consumers’ online consumption behaviour are determined, and the data characteristics are extracted by using classical TRA theory and decision tree; the similar feature data are determined by non-negative matrix decomposition method, the fused feature data are placed in two-dimensional space, and the noise points in the feature data are located by gradient matrix algorithm under Gaussian window. Determine the state of characteristic data, train the suspected abnormal behaviour data through the limit learning machine, randomly add weights and bias values in the training, output the results, and modify the results through the correction function to complete the anomaly identification. The results that the accuracy error of the proposed method is about 0.8%.
  Keywords: limit learning machine; online consumption behaviour; abnormal behaviour identification; TRA theory; Gaussian window.
  DOI: 10.1504/IJWBC.2023.10048829
   
  
• Decision making method of e-consumption behavior and attitude based on social network trust model  
  by Lan Qian 
  Abstract: In order to improve the problems of low accuracy and long time-consuming in the traditional decision-making methods of electronic consumption behaviour and attitude, this paper proposes a new decision-making method of electronic consumption behaviour and attitude based on social network trust model. Firstly, the social network trust model is used to calculate the comprehensive trust of e-consumption; Secondly, penalty clustering is introduced to divide the electronic consumption behaviour data, and the classification results of electronic consumption behaviour and attitude are obtained. Finally, the decision-making function of e-consumption behaviour and attitude is constructed, and the output result of the function is the decision-making result of e-consumption behaviour and attitude. The experimental results show that compared with the traditional decision-making methods, the calculation accuracy of the comprehensive trust value of this method is higher; it can accurately divide the consumer behaviour and attitude, and further shorten the decision-making time.
  Keywords: social network trust model; electronic consumption behaviour; attitude decision.
  DOI: 10.1504/IJWBC.2023.10048830
   
  
• Study on Network marketing service resource allocation based on social media platform  
  by Miao Wei 
  Abstract: To overcome the problems of low utilisation rate of network marketing service resources, high loss rate of resources and poor balance of resource allocation in traditional methods, a network marketing service resource allocation method based on social media platform is proposed. Build a social media platform for network marketing service resource allocation, and realise the two-way transmission of resources through the platform. SVM method is adopted to filter invalid and insecure data in social media platform. The PSO algorithm is used to configure the filtered service resources, and the cooperative game method is used to optimise the network marketing service resources, so as to realise the effective allocation of network marketing service resources. Simulation results show that the overall resource utilisation rate of the proposed method is not less than 90%, the maximum resource loss rate is only 1.4%, and the resource allocation is well balanced.
  Keywords: social media platform; network marketing; resource allocation; PSO algorithm; game theory.
  DOI: 10.1504/IJWBC.2023.10048831
   
  
• An online social network image retrieval using deep belief network  
  by Chao Guo, Hongzheng Dong 
  Abstract: In order to solve the problems of high retrieval error rate and long retrieval time existing in traditional online social network image retrieval methods, this paper proposes an online social network image retrieval method based on deep belief network. The restricted Boltzmann mechanism is used to build the deep belief network model, and the model is used to extract the image features of online social network. Cosine similarity calculation method is used to estimate the similarity of image feature vector, and online social network image retrieval is carried out according to the results of online social network image tag extraction. Experimental results show that the accuracy of online social network image feature extraction is always above 95%, and the error rate of image retrieval is between -1% and 2%, the average retrieval time of online social network image is 0.69s, and the practical application effect is better.
  Keywords: deep belief network; online social network; image retrieval; cosine similarity; image tags.
  DOI: 10.1504/IJWBC.2023.10049535
   
  
• Research on purchasing behavior prediction of e-commerce platform users based on multidimensional data mining  
  by Juan Long 
  Abstract: In order to overcome the problems of low accuracy of data mining, high relative error of prediction and high time consumption of traditional methods, a purchasing behaviour prediction method of e-commerce platform users based on multidimensional data mining is proposed. FP-Growth algorithm is used to mine the multidimensional data of e-commerce platform users’ purchasing behaviour, so as to extract the characteristics of users’ purchasing behaviour. Combined with feature extraction results, the hidden Markov model is used to calculate the probability of user access and the probability of no access. If the probability of access is greater than the probability of no access, it is considered that the user will access in the next month, otherwise it will not. The experimental results show that the average data mining accuracy of this method is 96.39%, the maximum prediction relative error rate is 5%, and the time consumption is always below 0.8 s.
  Keywords: multidimensional data mining; e-commerce platform; user purchase behaviour; behaviour prediction; hidden Markov model.
  DOI: 10.1504/IJWBC.2023.10049536