| Abstract: | 模式整合(Model Integration)為智慧型決策支援系統(Intelligent Decision Support Systems)的重要功能。到目前為止,雖已有許多模式管理(Model Management)的架構被提出討論,但仍缺乏能被普遍接受的自動化模式整合方法。在人工智慧技術中,黑板架構(Blackboard Architecture)提供隨機推論(Opportunistic Reasoning)及知識整合(Knowledge Integration)的特性。這種技術已被廣泛應用於科學及工程方面的專家系統中,其被運用於決策支援系統的模式形成(Model Formulation)方面之研究,也已有了起步。在決策支援系統中,模式形式強調新模式的產生,模式整合強調即有模式元件的重組及再用。本計畫以黑板架構為基礎建立一個模式整合方式,以黑板資料區(Blackboard Data Structure)、知識源(Knowledge Sources)、及控制(Control)三元件的適當展現方式,進行管理決策中資料對映及數量轉換過程,逐步完成模式整合功能。其分析動態則為執行「分析查詢-輸入資料-執行推論顯示結果」的循環動作,以因應使用者的查詢需求,提供重要的決策資訊供管理者參考。管理問題多半屬於半結構性或非結構性,在模式整合過程中,良好的解釋應為智慧型決策系統之重要功能。運用黑板架構的階層化問題表現方式、模組化知識源分類、及彈性控制之特性,本計畫亦嘗試以黑板架構建立提供完善解釋功能之智慧型決策支援系統,以讓使用者了解模式整合過程之問題點、知識源、及推理流程等。本計畫以兩個銀行管理方面的應用---資產負債管理及授信審查---為例,說明運用黑板架構於管理模式整合的具體作法,並分別探討其模式整合機能及解釋功能。
Model integration is an important aspect of intelligent decision support systems. Although several model management frameworks have been developed in the previous research, a widely accepted approach to automated model integration is still not available. Blackboard architecture is an AI technique which supports the opportunistic reasoning and knowledge integration by providing the structurization of problem solving space, modularization of domain knowledge, and flexibility of reasoning control. This technique has been used in various scientific and engineering problem solving domains. The research in applying blackboard architecture to automated model formulation in the managerial domain has also been attempted. In this research, we emphasize more on model integration which is defined as a series of processes for automatic data model mapping and mathematical model conversion to infer the important decision making information. We use the blackboard architecture to address the analysis dynamics in model integration in response to the user request for decision support. Since most managerial consultation problems are unstructured or semi-structured, good explanation facility should be able to enhance the consultation effectiveness. In this research, we also develop an explanation facility on the blackboard architecture to help the user understand the current problem points, the working knowledge sources, and the solution path. Using two bank management examples, Asset Liability Management and Loan Evaluation, we address the model integration mechanism and explanation functions individually. The prototype systems are developed to illustrate the concepts. |
