Summary: | 碩士 === 輔仁大學 === 資訊管理學系 === 90 === The technologies applied to automatic personalization and recommendation systems have become critical tools on the Internet environment, because they can help to cope with the problems of information overloading. In this thesis, we focus on integrating web usage and content mining to make personalized recommendation on news websites. The system would treat each user as an anonymous individual and identify user needs per session on web pages.
The quality of recommendation totally depends upon the profiles within the system. In this thesis, we propose a mechanism, News Concepts Indexing (NCI), hoping to extract concepts from news information in real time and produce content profiles. User access patterns are discovered by applying the techniques of association rules to the web logs and result in usage profiles. Our system will integrate content profiles and usage profiles to produce the user profiles for anonymous users per session on web pages. Based on the user profiles, the system not only understands what the user needs and recommends such news information, but also guides users to read news in a sequential manner.
Our experimental results, performed on real data, demonstrate that NCI can discover features of the concepts correctly, and the integration of web usage and content mining can increase accuracy of the resulting recommendation.
|