SAFPRS: Novel Framework of Sentiment Analysis for Lifestyle Product Recommendation System

Sentiment analysis is a significant tool for evaluating opinions in e-commerce applications since it is used to categorize customer feedback according to positive, neutral, and negative sentiments. These feelings are very useful in helping future customers to make the right decisions when shopping. This paper introduces a novel framework that employs two distinct deep learning-based approaches: For this purpose, we divide the general sentiment analysis into two categories which include the Sentiment Analysis with Enhanced Features (SAEF) and the Sentiment Analysis with Filtered Features (SAFF). Both models classify product reviews from which the sentiment polarity (positive, negative, or neutral) is derived and used for a recommendation system. These models work under the given framework and when the same dataset is used, separate customized recommendations for specific products are produced, emphasizing the fact that the framework outperforms the traditional methods. A key novelty of the proposed approach is that both models exhibit over 97% accuracy and this clearly shows how the approach improves on the recommendations and benefits the users greatly.