An Item-Oriented Collaborative Filtering Algorithm for Recommender Systems

Abstract

Recommender systems (RSs) are one of the popular, common and widespread applications in today’s business. It predicts the rating that a user is likely to give an item and finds a list of items to recommend the user. In this context, it is primarily categorized into collaborative filtering-based (CF) RS and content-based RS. In the CF, the RS suggests the items that are based on the behavior of similar users and its past behavior. One of the well-known algorithms in the CF is k-nearest neighbor (kNN), which relies on the behavior (rating) of k similar users (items) in oriented to users (items). However, it suffers from the unknown rating problem and the insufficient neighbor problem if the k similar users (items) have not rated the corresponding item (with respect to the corresponding user). Therefore, we develop an item-oriented CF (ICF) algorithm for RSs to overcome the problems associated with the kNN algorithm. The proposed algorithm ICF selects k similar items of a given item by finding the similarity count of the neighbors. We implement both the algorithms and compare with the kNN algorithm using four datasets to show the superiority of the proposed algorithm ICF in terms of two errors, namely mean absolute error (MAE), root mean square error (RMSE), precision, recall and F-score.