E-Commerce Recommender System based on Customer Reviews Analysis using Deep Neural Network

In recent days, users' preference for purchasing products through the web is increased drastically. Users can refer the recommendation about particular products based on other users' experiences on the web in various forms such as comments, blogs, reviews, etc. Before buying a product online, users read the review about the products which is based on the experience of others. A few people may leave reviews to boost the product's offer or to withdraw from it. Customers who trust reviews to decide whether or not to buy a product may be confused by this. As a result, it is necessary to find honest reviews and remove fraudulent reviews that have been added by malicious or fraudulent users. The proposed system comes up with a solution to this problem. The time interval between the reviews was calculated using driving times. The proposed system used Recurrent Neural Network (RNN) to precisely discover progressive fraud; the suggested approach mines dynamic periods, such as driving meetings. These drive meetings can be useful for discovering local inconsistencies in product reviews rather than global abnormalities. After that, examine the product's rating, reviews, and progression to identify evaluation-based facts, review-based facts, and chain of command facts. In addition, this work employs an optimization-based aggregation technique for coordinating real-world components in fraud detection. The evaluations of this optimization are based on obtained engineered Amazon product review dataset. The classification of product review data gives online users the information they need to understand review content in a short amount of time.