| 摘要 |
<P>PROBLEM TO BE SOLVED: To recommend an item on the basis of potential features of items which are not grasped completely in existent meta-information. <P>SOLUTION: On the basis of an assumption that "an item that an agent may visit is generated, after a certain substantial topic representing features of the item which are not grasped completely in known meta-information, in accordance with that substantial topic and an item probability distribution specific for a behavior zone of the agent guided from the known meta-information", a probability value θ (eigen-user topic appearance probability) for the agent to be interested in the substantial topic and a probability value ϕ (eigen-topic item selection probability) for a certain item to be selected from the substantial topic are calculated from a behavior history of the agent and item information, and the values are used to determine an item of high probability to be selected by the agent. <P>COPYRIGHT: (C)2013,JPO&INPIT |
