DecisionModelingofCPAusingVPAandEvolutionCBR

[[abstract]]財務報表分析對財報專家而言，是項耗時費力的非結構性工作，成本亦高，其分析成果有賴於專家的專業知識與多年經驗。但人為的決策會受到體力與精神的影響，難免產生倦怠和疲累，若能以資訊技術建構專家的決策模式，不但可以輔助專家的決策，提高其決策品質，也可以節省專家分析的的時間與精神。建立專家決策模式的主要方法有二：過程追溯法（process tracing）與結果分析法（output analysis），兩種方法各有其優缺點存在。因此，本研究針對個別財報專家的經驗與主觀意識，以財務報表分析中的短期償債能力分析為目標，在過程追溯法中採用口語草稿分析，在結果分析法中採用演化式案例推理，據以建構專家的決策模式。 我們使用口語草稿分析法，發展個別財報專家的初始決策模式，並請專家依其決策模式，針對746家上市上櫃公司的財務資料，分析其短期償債能力。我們收集專家分析後的案例，以案例式推理法結合遺傳演算法，驗證專家的決策模式。研究發現其結果大致與專家決策一致，不同之處亦可提供專家重新檢視其決策策略的機會，適時地給予專家適當的回饋。[[abstract]]Financial statement analysis is important in helping both management and stockholders understand the performance and future prospects of a company. The task of analysis requires specialized knowledge and long time experience to do the job well. Short-term liquidity is one of the major indicators in measuring the performance of a company. It takes experienced accountant to determine the quality of short-term liquidity of a company. This research attempts to construct a decision model based on human expert’s historical experience, so that the prediction capability of human expert can be repeated. In general, there are two methods in building a decision model of experts: process tracing and output analysis. Each of them has their advantages and disadvantages. We applied verbal protocol analysis in process tracing and evolution case-based reasoning in output analysis. Through verbal protocol analysis, we collected 746 cases and built a decision model, then, we integrated case-based reasoning with genetic algorithm to verify the decision model. We found, through experimental results, that the prediction capabilities of the model are in consistence with that of the expert