Recently, since variety of information through the web is excessively expanding, people are not only depending on searching engines, but also, spending more time for searching they need. In order to improve efficiency of the searching engine, recently, intelligent agents based on preference of each user are being developed. However, some intelligent agent depending on the very limited assumptions for the considering users preference produces almost irrelevant results.
Therefore, the intelligent agent is necessary to support a personalized searching function for the personal and active evaluation of the users preference. It also should support semantically related information by analyzing correlation of the information.
In this thesis, association rules and case-based reasoning method such that can analyze actions and characteristics of each user and extract individual behavior pattern rules by repetitive learning for supporting reasoning are suggested. The association rules and case-based reasoning method used in the data mining field can measure the degree of verification of the related data depending on the supportability and reliability, and can evaluate the degree of the relevance by its evaluation. Therefore, the recall ratio and the precision ratio can be improved by applying these rules for verifying the relevance among web documents.
The grouping rule formation algorithm for the relevance rule searching method and a recognized probability model for the implementation of the case-based reasoning method is presented. The former calculates supportability and reliability to measure the verification ratio, and the latter evaluates relevance by measuring the relativity.
The learning through the user modeling properly feedbacks the cleaned information of case presented level, and have expanded and adaptive function of the knowledge by persistently changing the users category group.
Moreover, in the study on the accessibility of the wired internet information, the advantages of improving the accessibility and the mobility can be verified by applying WAP technology on the web-searching agent. As a result of the performance evaluation, the association rules method showed better recall ratio than the expert searching engine and showed higher precision ratio than the general one. The case-based searching method showed better precision than both in the expert and the general one.
Therefore, the suggested searching engine using the association rules and the case-based reasoning method statistically based on the users searching behavior showed superior results both on the recall ratio and the precision ratio compare to the any other existing searching engines.