Scholarlink

<p>The main goal of this article is to devise an effective method named Cosine Migration Optimization-based Graph Neural Network (CMO_GNN) for product recommendation. Initially, the input data, such as user buying sequence, product id, name of the product, reviewer ID, and user review, is considered. Then, graph generation is performed, where the sequence is encoded. Later, a Graph Neural Network (GNN) is utilized to identify the relevant user by training it based on the user graph. The training of GNN is done on the proposed hybrid Cosine Migration Optimization (CMO). Then, product recommendation is performed by using the user’s buying behavior query as input to the GNN, which performs node classification to identify the relevant user. Later, the product buying sequence of the particular user is tracked from the user feature vector. Finally, the product recommendation is refined by analyzing the product’s sentiment using a trained sentiment classification model. Here, sentiment analysis is performed by considering the product review data. Subsequently, feature extraction is performed, and sentiment classification is performed by a Hierarchical Attention Network (HAN) trained by the CMO. Experimental results demonstrate that the proposed CMO_GNN achieves superior performance, attaining a precision of 90.995%, a recall of 91.945%, and an F-measure of 91.468%.</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

Sentiment Analysis for Product Recommendation Using Graph Neural Network with Cosine Migration Optimization

  • 期刊论文

  • 发表日期：

    2026年4月6日

查看全文
  • Sangeetha M,
  • R. Manjula Devi,
  • Lalitha Krishnasamy,
  • Kumaravel T

摘要

The main goal of this article is to devise an effective method named Cosine Migration Optimization-based Graph Neural Network (CMO_GNN) for product recommendation. Initially, the input data, such as user buying sequence, product id, name of the product, reviewer ID, and user review, is considered. Then, graph generation is performed, where the sequence is encoded. Later, a Graph Neural Network (GNN) is utilized to identify the relevant user by training it based on the user graph. The training of GNN is done on the proposed hybrid Cosine Migration Optimization (CMO). Then, product recommendation is performed by using the user’s buying behavior query as input to the GNN, which performs node classification to identify the relevant user. Later, the product buying sequence of the particular user is tracked from the user feature vector. Finally, the product recommendation is refined by analyzing the product’s sentiment using a trained sentiment classification model. Here, sentiment analysis is performed by considering the product review data. Subsequently, feature extraction is performed, and sentiment classification is performed by a Hierarchical Attention Network (HAN) trained by the CMO. Experimental results demonstrate that the proposed CMO_GNN achieves superior performance, attaining a precision of 90.995%, a recall of 91.945%, and an F-measure of 91.468%.