Abstract:
In online markets or shops, the products are purchased by people based on their visual appearance and textual description. These ecommerce systems also collect user views...Show MoreMetadata
Abstract:
In online markets or shops, the products are purchased by people based on their visual appearance and textual description. These ecommerce systems also collect user views or feedback about the sold product or services. The user also submits his complaints, reaction, and opinion including his sentiments. These reviews can be processed by ecommerce sites to rank their products and identify the features and flaws of products. In this paper, a weighted cause-reward analysis-based reinforcement learning method is proposed to optimize the sentiment classification. The model first transformed the input reviews to the weighted features by identifying the frequency and boundness with product. These weighted features are processed by the reinforcement learning method to classify the sentiments. The model is applied on the reviews collected from amazon reviews dataset. The analysis results are compared against machine learning and deep learning methods. The results shows that the proposed model achieved the average accuracy rate of 94.60% which is significantly better than existing methods. The performance and reliability of the system are improved by the proposed model.
Published in: 2022 International Conference on Smart Generation Computing, Communication and Networking (SMART GENCON)
Date of Conference: 23-25 December 2022
Date Added to IEEE Xplore: 06 April 2023
ISBN Information: