Item Based Collaborative Filtering Approach in Movie Recommendation System Using Different Similarity Measures

Abstract

Recommender systems use the concept of data analysis and knowledge discovery to essentially provide personalized recommendations on the web. A number of different algorithms has been proposed to effectively capture users interest and provide accurate recommendations. One of such promising technique is collaborative filtering. The predominantly used technique to date in recommendation system is collaborative filtering because of its effectiveness. User-based collaborative filtering approach is popularly and extensively used in practice but yet faces some key challenges in providing enough scalable and quality recommendations due to the daily increase of items and visitors in different websites. In this project we used collaborative item based technique to provide recommendations of items in real time by first analyzing the user-item rating matrix to determine the similarities between items in two different methods of computing items similarity (i.e. item-item correlation and item-item adjusted cosine similarity) and then use this similarity to predict unknown ratings of items, from which the top most rated items are chosen and be recommended to the intended user.