¡@

Information filtering is an area of research that develops tools for discriminating between relevant and irrelevant information. Users first give descriptions about what they need, i.e., user profiles represented by a set of keywords, to start the services. A profile index is built on these profiles. Then, the Web page will be recommended to the users whose profiles belong to the filtered results. Therefore, a critical issue of the information filtering service is how to index the user profiles for an efficient matching process. Among previous proposed methods, Wu and Chen¡¦s graph-based index method can expect to minimize the storage space. However, when the users often change their interests, the index structure of Wu and Chen¡¦s method needs to be reconstructed, resulting in the high update cost. Therefore, in this paper, we propose a data mining-based method for the incremental update of the index structure, the updatable tree, to reduce the update cost. In fact, each keyword could have a weight representing the degree of importance to a user. We apply this feature to distinguish between long-term and short-term interests. By making use of the property that the short-term interest has a higher probability to be changed than the long-term one, our proposed method can locally update the short-term interest, resulting in the low update cost. According to our experimental results, our method really can reduce the update cost as needed by Wu and Chen¡¦s method.

Keywords: data mining, incremental update, information filtering, personalization, profile

Retrieve PDF document (200801_09.pdf)

Received February 2, 2007; accepted July 13, 2007.
Communicated by K. Robert Lai, Yu-Chee Tseng and Shu-Yuan Chen.
^*This research was supported in part by the National Science Council of Taiwan, R.O.C. under grant No. NSC 95-2221-E-110-101 and by National Sun Yat-Sen University.