"Collaborative filtering" (CF) methods provide a good solution for recommendation systems. One of the main phases in CF is the neighborhood selection phase. It relies on selecting users according to their similarity to the active user. Unfortunately, almost all used similarity measures do not take into account many useful parameters associated with the users that can help computing similarity more accurately. This paper presents a comparative study of adjustment similarity measures that

more »

... Pearson correlation with various set-similarity measures (such as Jaccard similarity) as a correction coefficient. The focus is to improve computing similarity phase among users (items) to reflect as much as possible their real relationships. Finally, experiments using FilmTrust dataset show that combing Jaccard coefficient with Pearson similarity give more predictions accuracy than the traditional collaborative filtering.