New similarity for user-based collaborative filtering recommendation systems

Finding similarities between users is one of the most important factors in user-based collaborative filtering recommender systems that have important effects on the correct prediction of active user ratings. Until now, this similarity was calculated by special formulas such as Pearson, Tanimmota, Cosine, etc. We calculate it by proposing two similarity formulas with minimal calculations, and practical experiments show that the predictions of the obtained ratings are more accurate. In this method, the nearest neighbour is the one who has more common ratings with the active user

Keywords:

rating matrix , recommender systems , neighbor user