案例式規劃支援系統案例尋回機制之研究－－以工商綜合區為例

CASE RECALLING MECHANISM IN CASE-BASED PLANNING SUPPORT SYSTEMS-A CASE STUDY OF INDUSTRIAL-COMMERCIAL PARK PLANNING

案例式規劃支援系統，乃運用先前的規劃經驗來作為目前規劃之輔助。系統運作過程中，如何能從案例庫中找出合適的參考案例，此即所謂「案例尋回」之機制，其直接影響系統支援規劃能力，為系統建構的重要關鍵要素之一。然而一個完整的案例尋回過程，其涉及尋回案例的索引建立、案例選擇行為，以及不同案例間相似性的判斷方式等關鍵課題。本文嘗試以工商綜合區規劃為例，對於案例式系統建構時案例尋回關鍵課題加以探討。在尋回機制之索引系統建構部分，文中乃運用個人建構理論為基礎的角色建構測驗法加以修正後，建構包含12個索引項目與39個索引內容之索引系統。在規劃者選擇案例的索引使用行為方面，則透過調查掌握規劃者實際使用案例行為，加以判斷選擇案例的主建構與次建構之索引內容，以及各索引具有之補償與非補償等特性。而在規劃案例相似性判定方面，本文則針對不同資料尺度之索引，提出一套集合相似判定演算方法。最後，則運用Delphi語言整合地理資訊系統軟體（Arc-View 3.2）開發雛形系統，並透過「回收率」與「精確率」2個評估準則，對於實驗性系統運作進行初步檢核，並用以檢證系統之發展理論與發展方法。

A case-based planning support system applies previous planning experiences to assist the current planning. The system heavily relies on the 'case recall,' which is to seek appropriate references. The procedure of 'case recall' determines the success of the system. Therefore, 'case recall' is a key component of the case-based planning support system. The key topics include establishing the index of case recall, selecting usage of referenced cases, and judging among similar case in the process of whole case recall. The purpose of this paper is to use the industrial-commercial park planning to discuss the 'case recall' issues of the case-based system construction. In the case-recalling mechanism, the index system constitutes of 12 index items and 39 index contents, revised by the role construct repertory test (Rep Test), which is based on personal construct theory. Second, this research surveys potential users' selection behaviors to construct super-ordinate construct and sub-ordinate construct. Third, to select reference cases with similarity, this research also discusses the development of case set similarities and proposes a set of development of judgment of similarities from the different data scale indices. Finally, the research develops a prototyping system, which combines Arc View 3.2 (a geographic information system package) and a CBR shell (developed by the authors with Delphi language), relies on two criteria: 'recall ratio' and 'precision ratio' to evaluate the theorem and development method.