COLLABORATIVE FILTERING RECOMMENDER SYSTEM BASED ON SIMILARITY MEASURES USING THE ORIGIN MOMENT OF DIFFERENCE RANDOM VARIABLE

摘要

PURPOSE: A collaborative filtering recommendation system and a similarity table construction method based on similarity measure using origin moment of a difference random variable are provided to improve preference prediction efficiency by supplementing a problem generating a cold start condition when predicting a preference of a user. CONSTITUTION: A preference data receiver(1) receives preference about an item. A performance measurer(2) calculates an error of an item preference received from user and a preference anticipated in past and measures the performance of a system. A similarity measurer measures a user similarity and/or item similarity by using a similarity method of measurement.