Abstract:
Amazon has been the most dominant Ecommerce platform when compared with its competitors such as Walmart and eBay. A vital aspect of the products being sold within such platforms is a review, in which a customer review is a form of expression from customers, expressing their thoughts and experiences with the products and the services. One of the main review inconsistencies is the discrepancy between the review comment and the review star rating, and this research focuses on identifying such discrepancy occurring reviews of Amazon using a novel technique. Past research has made use of various techniques to identify such inconsistent reviews. Like Machine Learning models, Comparison techniques, Conditional Techniques and Correlation techniques. The author evaluated DNN, CNN and LSTM, and choose the best performing model, and after optimization, the LSTM model was chosen, which performed with an accuracy of 99% and it was integrated to a web application, introducing the prototype ‘Discrepancy Detector’ to detect inconsistent reviews.
