Improving Collaborative Filtering by Selecting an Effective User Neighborhood for Recommender Systems

In the last two decades, the increase in the amount of information available online have resulted in an information overload problem making it very complex for users to get the useful information they require within time. A recommender system helps customer to make useful decisions about products they wants to purchase thus providing better customer satisfaction which is vital in online environments such as e-commerce systems. Collaborative Filtering (CF) is one of the most significantly used method for generating recommendations for users. Collaborative Filtering depends on selecting a subset of users called user neighborhood for filtering recommendations for current user. In this paper, we have presented an algorithm that uses collaborative filtering for providing recommendations to a user by selecting an optimal neighborhood. The experiments are conducted by selecting two different neighborhoods; k-nearest neighbors and threshold-based neighbors. Results are generated by selecting different number of neighbors and selecting different threshold values for neighbors and generating recommendations. The results are compared in order to get the neighborhood which give the least error and better recommendation quality than others. The experiments are carried out using MovieLens data set with 1M ratings.