Comparative study of recommender systems built using various methods of collaborative filtering algorithm

Recommender system (RS) focuses on making personalized recommendations of items to the users by using the collaborative filtering (CF), which helps users to select appropriate items from the large dataset. CF is the method of predicting the usage of an item for any user, based on that particular user's interest that was shown previously or by taking opinion from other users. The purpose of RS is to provide personalized assistance to the users, for finding out the best items from the most used ones. The RS is also used to find the best items which have maximum popularity. There are some challenges in the RS due to the enormous growth of products and users over internet. The greatest challenge is to avoid data sparsity, cold-start and scalability problem to produce high quality recommendation. A Recommender System (RS) was built using the algorithms for combining the results of user and item based collaborative filtering technique by using Mahout and without using Mahout. The comparative study is performed using accuracy metric like F1 Score. Recommender system built using Mahout increases the scalability and quality for the predictions made and also it increases the recommendations of items to the users. “Movielens” dataset is used for the experimentation purpose of this paper.