The Impact of the k-Nearest Neighbor Parameters in Collaborative Filtering Recommender Systems

Recommender Systems (RS) have become very important recently, they are a main component of many applications in different fields. They aim to give beneficial information according to the profile of each user among the huge existing online information. RS are based on several approaches to provide the best results and give satisfaction to the active user. Collaborative filtering is one of these approaches. It helps to choose a product according to the consumer’s preference from many and various choices. It uses the k-Nearest Neighbor (kNN) technique for the extraction of similar users from the group of users. In this paper, we will study the effect of the parameters of the kNN algorithm on the obtained results. For that, we have varied the value of k, then we have measured for each value the prediction accuracy, using the Root Mean Square Error (RMSE), and the Mean Absolute Error (MAE) metrics. The experiments are carried out also to find the value of k which gives good results in both metrics. Then we have calculated also the RMSE and MAE metrics for different similarities in order to find the similarity which gives good results compared to others. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.