Performance analysis of recommendation system based on collaborative filtering and demographics

A recommendation system can personalize a website for a user by suggesting items that are according to the user´s liking. Recommendation systems predict the rating that a user would give to an item and recommend the top N items to the user. Collaborative Filtering algorithms are a class of knowledge discovery algorithms that give recommendations to a user based on the ratings of other similar users in the system. Traditional collaborative filtering algorithms face issues such as scalability, sparsity and cold start. In the proposed system, prediction using item based collaborative filtering is combined with prediction using demographics based user clusters in a weighted scheme. The proposed solution is scalable while successfully addressing user cold start. A performance analysis of the proposed system shows that it achieves a lower MAE and higher coverage than traditional collaborative filtering algorithms.