Recommendation System in E-Commerce Websites: A Graph Based Approached

Abstract

This paper is regarding the lack of semantic factor in recommendation systems and describes the different recommendation techniques that are being employed in the current e-commerce website. Recommendation system can be broadly classified into three categories: content-based, collaborative, and hybrid recommendation approaches. Content based systems consider the properties of the items to be recommended. For instance, if a Amazon user has purchased many romantic novels, then content based recommendation system recommends novels in the database as having the "romantic" genre. Collaborative filtering systems recommend items based on similarity measures between like minded users and/or items. The items recommended to a user are those preferred by similar users. This paper also emphasizes the need for semantics in current recommendation system to recommend products accurately. This also describes various limitations that are present in the current recommendation methods and suggests possible solutions that can improve current recommendation system used in e-commerce websites. It also includes a survey on popular e-commerce websites such as Amazon, Ebay, Flipkart Snapdeal and Paytm by rating them on different parameters and doing their comparative analyses This paper also focuses on how graph algorithm can be used to improve recommendation in ecommerce websites. The proposed system compares flickr.com recommendation of images with the proposed method. The method incorporates semantic recommendation using overlap technique based in graph.