Amazon Product Recommendation System

Abstract: This paper offers a detailed explanation of a system that uses sentiment analysis and machine learning algorithms to classify and recommend products on Amazon. Using the idea of Machine Learning, we developed a system that can be used by many e-commerce sites for better product recommendations. This system employs a machine learning model in which similar and superior products are offered to the customers in order of best to worst based on the product utilized in the past. The computer will compile a shortlist of all relevant items or products based on user-generated product reviews that meet the user's criteria, taking into account the product's quality and rating. The approach we employed was to create a system model that would analyze customer reviews for various products in the same category and then use Natural Language Processing to arrive at a conclusion where the system (model) would be able to assess whether the review is positive or negative. We've also used the ratings offered on various items to create a technique to combine ratings and reviews to improve the accuracy of the system (model). We employed the Collaborative Algorithm to improve the accuracy of product recommendations. During the creation of the system, we used the Amazon e-commerce site and its products to simulate a real-world implementation scenario (model). Our system uses cosine similarity to find the similarities between items on basis of the multiple user’s ratings and form a matrix which helps to recommend items to other users.

Keywords: Review & Ratings, Machine Learning, Natural Language Processing, Collaborative Algorithm, Recommendation System, Accuracy.