Genre based hybrid filtering for movie recommendation engine

摘要

With the dramatic rise of internet users in the last decade, there has been a massive rise in the number of daily web searches. This leads to a plethora of data available online, which is growing by the days. A recommendation engine leverages this massive amount of data by finding patterns of user behavior. Movie recommendation for users is one of the most prevalent implementations. Although it goes way back in the history of recommendation engines, collaborative filtering is still the most predominant method when it comes to the underlying technique implemented in recommendation engines. The main reasons behind that are its simplicity and flexibility. However, collaborative filtering has always suffered from the Cold-Start problem. When a new movie enters the rating platform, we do not have any user interaction for the movie. The foundation of collaborative filtering is based on the user-movie rating. In this paper, we have proposed a hybrid filtering to combat this problem using the genre labeled for a new movie. The proposed algorithm utilizes the nonlinear similarities among various movie genres and predicts the rating of a user for the new movie with the associated genres for the movie.