Neuro-fuzzy decision trees for content popularity model and multi-genre movie recommendation system over social network

In this paper, we propose a framework of multi-genre movie recommender system based on neuro-fuzzy decision tree (NFDT) methodology. The system is capable of recommending list of movies in descending order of preference in response to user queries and profiles. The system also takes care of attempt to vote stuffing using novel application of fuzzy c-means clustering algorithm. Typical user query and profiles consists of content ratings for multiple genres like action, comedy, drama, music and many others. The distinctive point of the proposed approach is to handle recommender system generation as a supervised pattern classification problem, where in user reviews for multiple genres are conditions and overall star ratings are decisions. The entire recommender system is represented in the form of NFDT. Rules represented by NFDT also acts as a tool for understanding the combinations of contents driving popularity (and unpopularity) over certain social network. We have also proposed a modified inference mechanism based on matching and ordering of firing strength of each fuzzy decision tree path in response to user queries. The computational experiments have been presented on a sample real-world movie review database to judge the efficiency of the proposed recommender system.