Using Integrated Approach To Build Financial Forecasting Model- Case With Electronics Industry In Taiwan After The Financial Crisis

[[abstract]]當公司財務有問題時，投資人與銀行為了減少損失，必需要有一套完整預警機制能預測公司未來的財務狀況，而為了建立完整的預警機制，必須由研究變數類型、特徵篩選方法、預警模型等三方面的配合，在本文裡，使用金融海嘯後的上市櫃電子業公司為研究樣本，再從中找出財務比率與非財務比率，加以分成二組，一組為僅使用財務比率，另一組為使用財務比率與非財務比率，再分別由逐步Logit模型與因素分析篩選變數後，輸入至Logit模型、類神經網路、支持向量機，形成多個的預警模式，最後以所有預警模式得出平均交叉驗證的預測準確率與錯誤分類成本，來檢視出最佳的預警模式。 When the company has financial problems, investors and banks will need to predict the future financial position of the company to reduce some loss of money by a complete set of early warning mechanism. An early warning mechanism must coordinate three aspects form study variable type, feature methods, early warning models. In this article, the study sample is listed companies in the electronics after the financial crisis, and it is found financial ratios and non-financial ratios, to be divided into two groups. one group use only financial ratios, another use financial ratios and non-financial ratios, then using stepwise logit model and factor analysis select variables. This selected variables input logit models, neural networks and support vector machines, to be composed of multiple warning modes. Finally, using the average cross-validation of prediction accuracy rates and misclassification costs examine all early warning modes to find the best early warning mode