Novel Approach using user based Similarity for Recommendation Systems

Recommendations Systems (RS) are widely used by E-commerce and other web sites to recommend items to its users. Collaborating Filtering (CF) is very successful technique used in many web sites for recommendation. Many hybrid techniques with collaborative filtering are proposed in recent years for increasing the performance of recommendation system. In this paper a new user similarity based algorithm is proposed for recommendation system. The algorithm is based upon similarity between the users. The proposed algorithm is applied on a sample data set and its performance is calculated. The value of Mean Absolute Error (MAE) for the proposed algorithm is very reasonable.