A Recommendation System Based On Sentiment Analysis

Abstract

The recommendation system is an essential component of multiple domains, such as e-commerce, health care and social media, where it predicts user preferences by providing relevant suggestions. In this work, to improve the performance of the recommender systems, we propose a new approach to recommendation that we have applied on the Yelp Business dataset. This approach is based on sentiment analysis using three Machine Learning (ML) methods: Support Vector Machines (SVM), Naive Bayes (NB), Logistic Regression (LR) and Bidirectional Encoder Representations from Transformers DistilBERT. The proposed approach aims to improve recommendations through sentiment analysis of their user’s reviews. We employ the method of Singular Value Decomposition (SVD) to help treat sparsity in user-item interaction data, ensuring minimum variance results with limited data inputs as well. The results using Neural Collaborative Filtering (NCF) as a recommendation model showed a good performance of the proposed approach, demonstrating its efficiency and accuracy in recommendation tasks.