A competitive information recommendation system and its rational recommendation method
A competitive information recommendation system and its rational recommendation method
复制标题
一种竞争性信息推荐系统及其理性推荐方法
DOI：
10.1002/scj.v38:9
复制
发表时间：
2007
期刊：
影响因子：
--
通讯作者：
S. Tatsumi
中科院分区：
文献类型：
--
作者：
Toshiki Sakamoto;Y. Kitamura;S. Tatsumi
检索引擎被广泛地用作用户从互联网上的海量多样化的Web信息中获取满足其需求的信息的一种手段。然而，除非用户的需求被明确地表示为关键字，否则检索引擎是不会有用的。因此，信息推荐系统作为一种提供用户可能感兴趣的信息的手段正引起人们的注意。本文提出了一种竞争信息推荐系统，通过多个具有个性的智能体与用户的竞争性交互来向用户推荐信息。在竞争信息推荐系统中，如果代理简单地自主推荐信息，则可能不推荐用户想要的信息，或者在用户获取想要的信息之前可能需要进行多次交互。因此，本文提出了一种合理的推荐方法，该方法用一个多属性效用函数来表示推荐信息对用户的效用，在学习每个属性的权重时，代理只推荐用户想要的信息。然后，给出了以主体利益为优先的最优利润策略和以用户的效用学习为优先的最优学习策略，作为对主体的信息推荐策略。评价实验表明，最优盈利策略效果更好。《威利期刊》杂志，38(9)：74-84,2007；在线发表于《威利国际科学》()。DOI 10.1002sscj.10662
Retrieval engines are widely used as a means for users to acquire information satisfying their requirements from the large amount of diversified Web information on the Internet. However, retrieval engines cannot be useful unless the requirements from the user are explicitly represented in terms of keywords. Consequently, information recommendation systems are attracting attention as a means of providing information which is likely to interest the user. This paper proposes a competitive information recommendation system, which recommends information to the user by competitive interaction of multiple agents, each having an individual character, with the user. In the competitive information recommendation system, if the agent simply recommends the information autonomously, information desired by the user may not be recommended, or many interactions may be needed before the user acquires the desired information. Therefore, this paper proposes a rational recommendation method in which the utility for the user that can be obtained from the recommended information is represented by a multi-attribute utility function, and the agent recommends only information desirable to the user while learning the weight for each attribute. Then the best profit strategy, in which the profit of the agent has priority, and the best learning strategy, in which learning of the utility to the user has priority, are presented as information recommendation strategies for the agent. An evaluation experiment shows that the best profit strategy is better. © 2007 Wiley Periodicals, Inc. Syst Comp Jpn, 38(9): 74–84, 2007; Published online in Wiley InterScience (). DOI 10.1002sscj.10662
DOI：
10.1613/jair.105
发表时间：
1994-08
期刊：
ArXiv
影响因子：
--
作者：
Thomas G. Dietterich;Ghulum Bakiri
通讯作者：
Thomas G. Dietterich;Ghulum Bakiri
一种针对假名投标的鲁棒公开升价多单位拍卖协议
DOI：
--
发表时间：
2005
期刊：
Decision Support Systems Vol.39(Refereed)
影响因子：
--
作者：
Atsushi;Iwasaki;Makoto;Yokoo;Kenji;Terada
通讯作者：
Terada