USER-BASED COLLABORATIVE FILTERING RECOMMENDER SYSTEM AND METHOD AMENDING SIMILARITY USING INFORMATION ENTROPY

摘要

PURPOSE: A user based collaborative filtering recommendation method for revising similarity by using information entropy and a system thereof are provided to revise the similarity of a recommendation object user through the information entropy of the recommendation object user and other user, thereby improving the accuracy of the preference prediction of the recommendation object user. CONSTITUTION: A similarity calculator(32) calculates similarity between recommendation object user A and other user B. An entropy calculator(34) calculates information entropy about each user. A similarity compensator(33) obtains a weighted value by the information entropy. The similarity compensator obtains compensated similarity by multiplication between the weighting value and the calculated similarity. A preference estimator(36) predicts the preference of the recommendation object user.