DECISION SUPPORT ALGORITHM FOR PREDICTION OF THE RATING OF PRODUCTS WITH NO RATINGS BASED ON THE MEASURE OF SIMILARITY OF THE STATISTICAL IMPLICATION

Vo Thi HuyenTrang, Kvyatkovskaya Irina Yu, Tran Quoc Toan

The measure of similarity plays an important role in the recommender system of collaborative filtering based on users, since it directly affects the results of recommender systems. To determine the measure of similarity between two users in a recommender system, many solutions are proposed, such as: using statistical correlation, using cosine distance between two vectors, using association rules, etc… In this paper, he proposes a measure of similarity based on the statistical implication analysis method. The measure of similarity between two users is determined based on the sum of the statistical implication distances of the association rules, which tend to favor counter-examples (the number of objects that satisfy property a but not property b in the association rule ) generated from the ranking data of the two users.

Keywords:similarity measure; recommender system; statistical implication analysis; statistical implication intensity measure; product rating; rating matrix; association rule