摘 要 企業營運狀況會定期反應在財務報表上,但卻需等到財務揭露之後,投資人才能預警公司是否出現狀況,也因為「知識經濟」時代的來臨。以智慧資本為主的無形資產已成為企業重要的核心競爭力,而許多財務及非財務資訊經先前學者證實可幫助企業預測財務危機的發生,但智慧資本資訊指標因素卻鮮少學者深入研究。 本研究將使用資料探勘的類神經網路、決策樹及支援向量機等人工智慧方式配合逐步回歸、因素分析等資料篩選方式來建立企業危機預警模式,以企業財務比率、非財務變數和智慧資本指標為輸入變數建構出企業危機預警模式。研究樣本選取26家電子產業之財務危機公司,以一比一配對方式選取26家電子產業相同類型相同資本額相近的正常公司,共52家公司樣本,資料蒐集使用32項財務比率和12項非財務指標變數及16項智慧資本指標作為輸入變數,來進行研究模型的建立。 在本研究實驗資料中發現,結合了財務比率指標、非財務指標變數和智慧資本指標能有效提升預警模型準確率,也優於先前學者的財務預警的準確率,此外,本研究實驗數據也顯示,以支援向量機所建立的預警模型的準確率勝過其他4種預測方法。
Abstract Financial statements reflect enterprise operating statuses. Investors can obtain complete information once formal financial statements are disclosed. However, if executives of firms intentionally embellish financial statements, investors cannot get the actual picture of the enterprise operation from the disclosed financial statements. With the advent of the era of “knowledge-based economy”, intangible assets, mainly relying on intelligence capital, have created competitive edges for enterprises. In previous literature, financial and non-financial information has been proved beneficial in predicting financial crises. Nevertheless, research applying intellectual capital indicators to foreshadow financial crises is still lacking. This research adopts neural networks, decision trees, support victor machine data mining techniques, as well as stepwise regression and factor analysis to establish a forecasting model. Additionally, it uses financial, non-financial, and intellectual capital indicators to predict corporate financial crises. 26 electronic corporations in financial crises and 26 corporations with stable financial status in the same industry have been chosen as samples. Furthermore, this research utilizes 32 financial indicators,12 non-financial indicators, and 16 intellectual capital indicators to construct a forecasting model of enterprise financial crisis. This research finds that the combination of financial, non-financial, and intellectual capital indicators has efficaciously enhanced the accuracy of the forecasting model, which is higher than that of the financial forecasting models established in previous studies. In addition, this research shows that the accuracy of the forecasting model based on the support vector machine surpasses the accuracy of other 4 prediction methods.