Collaborative Recommendation System Using Dynamic Content based Filtering, Association Rule Mining and Opinion Mining

This paper proposes a recommendation system (RS) that generates items recommendations to users with the help of dynamic content based filtering, collaborative filtering, association rules and opinion mining. This RS uses dynamic content based filtering for creating and continuously monitoring the changing shopping behaviour of users. The proposed approach finds other like-minded people with the target user that may cooperate with each other, in the form of items ratings using collaborative filtering. The approach uses association rule mining for the analysis of current market trend. It generates association rules only from those items that are liked by the users. Most of the people prefer to read reviews about the product, before purchasing. Almost all well-known e-commerce websites have hundreds of product reviews available, so it becomes very difficult for the user to read each and every review before buying any item. The proposed approach uses its own unique weighted opinion miner that summarizes the reviews and generates the weights for each item based on customers’ reviews. These weights help in estimating the popularity of an item among customers. This RS generates the final recommendations to users by combining the outputs from the collaborative-classifier, association rules and weighted opinion miner. The proposed RS is evaluated over live dataset using precision evaluation metric. The result shows that recommendations generated by proposed method out performed existing benchmark recommendations methods.