A similarity measure model based on the dissimilarity degree between users

Abstract

This article offers a new users-similarity measure to improve recommender systems' accuracy even when a few ratings are available. This model is based on the degree of dissimilarity between common ratings. On real data sets, we conducted our experiments to evaluate the proposed model's accuracy. Experiments demonstrate the model's capacity to locate a suitable neighborhood, resolve the user's cold-start problem, and significantly improve recommendation accuracy. Thus, this research may be considered a supplement to previous studies on user cold-start and good neighbor selection difficulties in user-based collaborative filtering algorithms.