近年資本市場頻傳企業財務危機而倒閉，致使許多投資人蒙受其害，也危害了資本市場的安定性，讓政府頭痛不已，因此財務危機預警模型是很好的預警防範工具，可以在企業發生財務危機之前就先偵測到並提前做出相關因應準備，將傷害降到最低。最早財務危機預警模型是由單變量分析，爾後又演變到多變量分析，近年來很多人工智慧演算法的蓬勃發展，使得更多更新穎的資料分類技術被發展出來如類神經網路、支援向量機等，這些新的人工智慧演算法技術不但突破於傳統統計上的限制，並且運算準確率更高，使得在企業界應用甚廣。因此，本研究嘗試以傳統的區別分析和遺傳基因演算法結合較新之人工智慧演算法「最小平方法支援向量機」來建立一個三階段的財務危機預警模型，本模型將比以往財務危機模型準確率更高、更準確。
Recently, there are a lot of enterprise bankruptcies due to financial distress. Huge numbers of investors suffer assets damage. It also damages the stability of the Capital Market, and causes great concern to the Government. Hence, the Financial Distress Prediction Model is a very useful tool to predicate the financial distress. It can detect the financial distress before it really happens and prepares the methods to deal with the financial distress. This will effectively decrease the damage to the Capital Market. The earliest predication model was made by Univariate Analysis. After that the model was developed by Multivariate Analysis. Today, the high development in Machine Learning produces a lot of methods like Neural Network, Support Vector Machine…etc. These new Machine learning methods not only surpass the limitations of Statistics but also increase the accuracy. They are applied widely in the business. Therefore, this study tries to combine the Discriminant Analysis and Genetic Algorithms with Least Squares Support Vector Machine to make a 3-level Financial Distress Predication Model. This model will be more powerful than the previous model.
