Scholarlink

<p>With the advancement of social media, the information diffusion popularity prediction has attracted wide attention in many applications. However, due to the real-time changes in networks and the complexity of social interactions, analyzing the exact mechanism of the information dissemination process remains extremely difficult. However, conventional popularity prediction methods rely heavily on human expertise to create features and define the generative model, or completely depend on the underlying user relation network for embedding learning. To address these concerns, this proposed work designed an information diffusion prediction model using a Hybrid Graph Convolutional Neural Network–Long Short Term Memory (GCNN–LSTM). The sentiment of the diffused information is analyzed through a hybrid Long Short Term Memory–Support Vector Machine (LSTM–SVM) in social media applications. The proposed work comprises two phases: the first phase is for effective popularity prediction of information diffusion, and the second phase is for analyzing the users' sentiment based on the influence of diffused information. In the detection of the information diffusion phase, the user data from social media is gathered to make a graphical representation based on the comment node attributes, and the effective features are extracted with the assistance of the graph convolutional neural network. After that, the features are given in the LSTM model for the detection of popularity. The diffused information in social media is considered in the sentimental analysis phase; initially, the data is subjected to preprocessing, and features are extracted with Term Frequency Inverse Document Frequency (TFIDF). Finally, with the help of a hybrid LSTM–SVM, the sentiment of the social media data is detected. The proposed model is implemented in MATLAB software and achieved 96% accuracy for phase 1 and 96% for phase 2. Thus, the proposed model effectively detects the spread of information in social media, which can assist various applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Information diffusion prediction using hybrid GCNN–LSTM and stock market based sentiment analysis in social media

  • 期刊论文

  • 发表日期：

    2025年6月25日

查看全文
  • Shweta Mayor Sabharwal,
  • Niyati Aggrawal

摘要

With the advancement of social media, the information diffusion popularity prediction has attracted wide attention in many applications. However, due to the real-time changes in networks and the complexity of social interactions, analyzing the exact mechanism of the information dissemination process remains extremely difficult. However, conventional popularity prediction methods rely heavily on human expertise to create features and define the generative model, or completely depend on the underlying user relation network for embedding learning. To address these concerns, this proposed work designed an information diffusion prediction model using a Hybrid Graph Convolutional Neural Network–Long Short Term Memory (GCNN–LSTM). The sentiment of the diffused information is analyzed through a hybrid Long Short Term Memory–Support Vector Machine (LSTM–SVM) in social media applications. The proposed work comprises two phases: the first phase is for effective popularity prediction of information diffusion, and the second phase is for analyzing the users' sentiment based on the influence of diffused information. In the detection of the information diffusion phase, the user data from social media is gathered to make a graphical representation based on the comment node attributes, and the effective features are extracted with the assistance of the graph convolutional neural network. After that, the features are given in the LSTM model for the detection of popularity. The diffused information in social media is considered in the sentimental analysis phase; initially, the data is subjected to preprocessing, and features are extracted with Term Frequency Inverse Document Frequency (TFIDF). Finally, with the help of a hybrid LSTM–SVM, the sentiment of the social media data is detected. The proposed model is implemented in MATLAB software and achieved 96% accuracy for phase 1 and 96% for phase 2. Thus, the proposed model effectively detects the spread of information in social media, which can assist various applications.