| 摘要 |
PROBLEM TO BE SOLVED: To reflect precisely the need of a user, and to extract a recommended commodity by a small processing load. SOLUTION: The number of times of various actions (browsing of detailed explanation, registration to a basket, actual purchase or the like) of each user relative to each commodity which is a sales object is registered in an action history database 17, classified into the recommended case of each commodity and the not-recommended case thereof. Scores of each commodity are determined by setting the point of the action more similar to the actual purchase to be higher and the action performed in the not-recommended case to be higher in the history of the registered actions, and registered in a commodity score database 18. When accessed from a user terminal 2, a CPU 11 extracts commodities having large scores registered in the commodity score database 18 as many as the prescribed number relative to the user, and presents them to the user as recommended commodities. |
