A User-Oriented Collaborative Filtering Algorithm for Recommender Systems

Abstract

Recommender systems (RSs) are one of the emerging applications in electronic commerce companies, such as Amazon, Flipkart, eBay, Levi's and many more. It generates a list of probable recommendations for the customers or users of the companies in one of the two categories, namely collaborative filtering-based (CF) and content-based. In CF, the recommendation is based on a user's past behavior and other similar users' behavior. Many algorithms have been developed for finding the recommendation in CF recommender systems. One of the popular algorithms is $k$ -nearest neighbor ( $k\text{NN}$ ), in which the recommendation depends on the behavior of $k$ similar users. More specifically, the user rating of a non-rated or unpurchased item is the aggregate value of $k$ similar users. However, the user rating is unknown iff the $k$ similar users have not rated the corresponding item. In this paper, we propose a user-oriented CF algorithm for RSs. The proposed algorithm selects $k$ similar users by finding the similarity count with other users for a given item. We implement the proposed algorithm and compare with $k\text{NN}$ algorithm. Simulations using four generated datasets show the supremacy of the proposed algorithm in terms of the mean absolute error (MAE), root mean square error (RMSE), precision, recall and $F$ -score.