Prediction of Product Rating based on Polarized Reviews using Supervised Machine Learning

Abstract

E-commerce websites facilitate customers to leave their experiences in the form of textual reviews for a variety of products. Recently, online reviews have played significant influencing role in customers’ decision for purchasing. The reviews have information and first hand experience about products’ quality for customers. Free-text sections are frequently found on online review web pages in addition to star-level reviewing options. But on many web pages, we find only the former option. Therefore, there is a need to convert the text-written reviews to star-level on the basis of the information they contain. Automatic conversion of online text-based reviews has recently been emerged as an active field of research in machine learning and deep learning. This paper presents a supervised machine and deep learning based solution to transform text-based reviews to star-level numerical representation by exploiting polarization detected on the basis of lexical analysis. Experiments were conducted on famous Amazon dataset under different choices of regression and classification techniques. Experimental results have indicated that the use of polarized reviews can significantly improve the rating prediction.