財務報表分析對投資者、債權人與企業管理者而言是一項重要的決策資訊來源，然其分析過程本身即為一項複雜且耗時的非結構性決策工作，其成效端視專家或專業會計師的決策策略;於此，決策支援系統或專家系統足以扮演以模式輔助專家的角色。然如何有效擷取專家知識以建立其決策模式，則為一艱鉅之任務。不同於以往相關之研究方向，本論文探討如何整合決策理論中之過程追溯法與結果分析法建構個別財報專家分析短期償債能力的決策模式，以期能改善其決策品質並達到以模式輔助其個別的判斷及決策的目的。研究人員首先應用口語草稿分析法擷取專家進行決策之認知過程，以問題行為圖表徵專家解決問題的步驟及使用線索。接著運用機率型類神經網路分析由專家以其所使用的線索進行解題後的國內165家上市公司案例資料，建立該專家之初始決策策略，並更進一步藉由深入分析專家與類神經網路之不一致處，回饋給該專家使之得以回顧並修正其決策模式。實證結果顯示，藉由本研究所提之整合模式法所建立的專家決策模式，確實可以提供專家檢視並修正其決策模式的機會，亦可降低一般直接採用過程追溯法之研究所可能發生的缺失。

The analysis of financial statement analysis is one of the major decisive information sources for investors, lenders, and managers. It is an unstructured decision problem in nature and a complicated and time-consuming process whose effectiveness is dependent on the decision strategy of expert or CPA. Recently decision support systems or expert systems have been developed for tackling with such work. However, it is of greatly importance and is a non-trivial job for building the decision model based on the knowledge extracted from experts. In this study, we integrate the methodologies of process tracing and output analysis in decision theory to construct the decision model for a senior professional CPA and to improve the decision quality of the expert. First, the recognition process of the investigated expert is identified by using verbal protocol analysis and is represented as prob­lem behavior graphs. Second, an initial decision model of the expert is built from 165 empirical cases he evaluated via applying probabilistic neural network (PNN), a non-blackbox model which allows one to look into the impact of independent variables. The expert is then asked to review and refine his decisive strategy according to the inconsistent cases between him and PNN. The empirical study findings show that can the decision model constructed by the methodology proposed not only alleviate the shortcoming resulting from directly appealing to neural networks but also provides a transparent investigation for decision-making process.