本研究使用財務比率及智慧資本建構Logit模型及倒傳遞類神經網路之預警模型,並以電子產業台灣上市櫃公司為研究對象。本研究以1999年2014年曾發生全額交割之上市櫃公司做為樣本,總共79家危機公司。再分為樣本內47家危機公司,樣本外32家危機公司。 根據實證結果發現,不論使用Logit模型及倒傳遞類神經網路之預警模型,預測期間離發生危機時點越短,預測準確率越高。Logit模型未加入智慧資本前四期,預測準確率之平均為79.30%;加入智慧資本前四期,預測準確率之平均為82.03%,因此前四期的平均準確率提升2.73%。倒傳遞類神經網路未加入智慧資本前四期平均為78.91%;加入智慧資本前四期平均為82.42%,因此前四期平均準確率提升3.51%。由此可知,加入智慧資本後確實有助於提升預測準確率。
In this study, the sample is from publicly traded companies of Taiwan's electronics industry. We use financial ratios and intellectual capital to construct a financial distress model by Logit model and back-propagation network. The data is full-cash delivery stock during 1999-2014 from total 79 companies which is in crisis. According to the empirical results, the forecast will more accurate, when the sample period is more close to the crisis point. No matter we use Logit model or back-propagation network, the result will keep in similar. After we adding the variable of intellectual capital to both model, the former four average forecast accuracy ratio raised 3.15% in Logit model model, and in the back-propagation network it raised 3.51%. So we suggest that, adding the intellectual capital may improve the prediction accurate.