A Collaborative Filtering Recommendation Algorithm Based On User Clustering And Item Clustering

Recommendations that are personalized help the users in getting the list of items that are of their interest in e-commerce sites. Majority of recommender systems use Collaborative Filtering techniques to generate recommendations to their users. This project implements an information filtering technique called as Collaborative Filtering for generating personalized recommendations in movies for user. Collaborative Filtering is of two types, namely, collaborative filtering based on users and collaborative filtering based on items. Collaborative Filtering based on users is more expensive computationally but it produces better results. Collaborative Filtering based on users is not preferred because it encounters the problems of Scalability when the number of users increases. Therefore, we use item-based Collaborative Filtering which is an alternative method. Collaborative Filtering, which is based on items uses two techniques-Pearson correlation technique and Adjusted cosine technique for calculating the similarity between items and to generate recommendations to users. In this Project both the above techniques are used and to measure the accuracy of the predictions generated by these techniques, Root Mean Square Error is computed. iii CONTENTS Contents Page No.