應用 RB 與 CBR建構一個解答擷取系統

Abstract

由於網路快速發展，服務多人上線的系統，已從傳統的兩層式 (client-server) 架構轉型成三層式 (client-application server-database)架構，因此，問題的偵測跟解問題的方法，對於這樣的領域來說，其複雜度就不斷增加。就我們所知，很多公司需要花很多人力物力去維護這樣的系統，然而公司的專家不可能一直在公司解問題，既使用文件搜尋來找答案，也可能找不到解答，所以提出根據專家系統的方法，來建構一個解答擷取系統。在本篇論文中，我們提出 Solution-Retrieval System (SRS) 架構來輔助使用者解決問題。在SRS中，結合 Rule Base (RB)跟 Case Based Reasoning (CBR)方法，我們使用RB推論來縮小問題的範圍，用CBR來找問題的解答，這樣處理的方式是模仿專家的處理模式，依著系統的特性，用推論的方式來找出錯誤類型，再將推論出來的錯誤類型，用專家的經驗法則，比較相似度，將相似度高的解決方案應用到問題上去解系統問題。再者，為了提高解問題的及時性，我們的系統也可以在 PDA等小型裝置上及時的運用來解問題。未來，這樣的架構可以應用在相關的領域上，例如: IC 設計，整合供應鍊 等。

Due to the fast development of web services, most of the service activities have been moved from 2 tiers (client-server) to 3 tiers (client-application server-database); hence, the importance of problem diagnosis and solution retrieving in integrated domains becomes more complicated. As we know, many of companies devote lots of time and effect to deal with this problem. However, since experts are not always available, using traditional approach or knowledge management center to search solutions may still fail. Hence, the idea of developing a solution-retrieval system based upon expert system approach is proposed. In the thesis, we propose an architecture based on rule base (RB) and case-based reasoning (CBR) and build a Solution-Retrieval System (SRS) to help users to solve the problems, in which RB is used to reduce error scope and CBR is used to find the corresponding solutions. Similar to the expert’s diagnosis approach, we use the SRS to diagnose the error type by RB inference, and retrieve solutions by CBR. Finally, the retrieved high similarity solution cases can be used to solve problems. Furthermore, the PDA and hand-held devices could be used in our system for solving problem promptly. In the near future, this architecture will be adapted on other related domains, e.g., IC design and Supply Chain.