| 摘要 |
PROBLEM TO BE SOLVED: To recommend even a commodity which is not purchased by anyone, with high accuracy, in an E-Commerce service.SOLUTION: A commodity recommendation device uses purchase information and meta data to estimate each purchase topic, uses purchase information and meta data to estimate a meta data topic that is relevancy of each element in meta data, uses purchase information, meta data, purchase topic, and meta data topic to estimate the size of influence degree of the meta data on user purchase trend, uses purchase information, meta data, purchase topic, a meta data topic, and influence degree to calculate a probability of each user purchasing each commodity, and stores it as a recommendation score in recommendation score storage means. If a user identifier is given, a recommendation score for each commodity is acquired, and higher-order N commodities in the recommendation score are recommended. |
