Recommendation system to accomplish user pursuit

Abstract

Recommendation system provides information about the arrival and importance of a newly released movie to their registered user. The pursuit of the users is analyzed from their past history. In this paper, a recommendation system is proposed to recommend rating of the movie to the users. The learning phase of the system takes in the user particulars about the user till-date and his rating towards those movies. Having the Genre of the movie and its rating, the system is trained by data mining classifiers like Bayesian, Multiclass Classifier, Decision Stump Tree, Best First Decision Tree(BFTree) and Radial Basis Function(RBF) and the classification parameters i.e. True Positive rates(TP), False Positive rates(FP), Precision, Recall and Mean Absolute Error are computed. It has been concluded that the RBF classifier performs better than the other classifiers. This paper also focuses to address the problem of cold start movie. The genre of the new release is obtained and it's recommended to the corresponding user, those who are closely correlated. Implementations are carried out using movie lens datasets.