Hybrid Movie Recommendation System Using Opinion Mining Approach

Abstract

Recommendation Systems are utilized to assist clients with dealing with the data over- burden issue by creating customized content and items as per their inclinations. Past the customary recommender methodologies, there is a developing exertion to fuse user reviews into the suggestion procedure, since they give a rich arrangement of data with respect to the two things ’ features and users’ preferences. Recommendation system utilizes the user opinion and their sentiments about the features (example, genres in movies) to give recommendations and produce output. It is observed in this paper that text based reviews from the user can be very optimal in improving the accuracy and efficiency of a movie recommendation system. Three approaches have been used to extract features from the movies aspects to be used as features for collaborative filtering. The required manual interactions are different for all the approaches. The dataset that was used for the evaluation of features with collaborative filtering comprised of ordinal (star) ratings of several thousand movies. An elaborative algorithm of collaborative filtering is applied for the purpose of recommendation during evaluation and to compare performance of the system with inclusion of features from user reviews and without the features from user reviews. The assessment mining based highlights perform altogether better that the basic methodology, which is simply based on ordinal ratings and movie features like movie genres.