應用資料探勘技術於財務危機預測之研究

企業發生財務危機會對國家的經濟造成重大影響，所以企業財務危機預測一直是一個非常重要的研究主題，企業財務危機預測可以幫助企業盡早找出發生財務危機的前兆，讓有可能發生財務危機的公司能夠及早發現，及早解決問題，及時將企業拉回正軌。先前學者們的研究通常是在變數中尋找造成財務危機的關鍵變數，或是建立特定產業的財務危機模型，在過去的研究發現在變數中加入公司治理變數能使模型準確率提高，故本研究在變數中加入了董監持股比率、董監質押比率及大股東持股比率三項公司治理變數。本研究嘗試使用將資料以總資產切分成前二分之一及後二分之一，並進行模型比較。本研究在模型方面使用了決策樹、類神經網路及隨機森林，資料方面使用TEJ經濟新報之2016年至2019年公司資料，進行準確率、型一誤差及型二誤差的比較，找出最佳的財務預警模型。

關鍵字

財務危機；決策樹；隨機森林；類神經網路

並列摘要

Enterprise financial crisis will have a great impact on the national economy, so enterprise financial crisis prediction has always been a very important research topic, enterprise financial crisis prediction can help enterprises to find the foremen of the financial crisis as soon as possible, so that enterprises with the possibility of financial crisis can find and solve problems early. Previous scholars' research is usually to find the key variables causing the financial crisis in the variables, or to establish the financial crisis model of a particular industry, in the past, the study found that adding corporate governance variables to the variables can improve the accuracy of the model, so this study has added three corporate governance variables in the variables, the ratio of the shareholding of the supervisor, the ratio of the pledge of the supervisor and the ratio of the majority shareholder. In this study, we tried to divide the data into the first and second parts of the total assets and compare the models. In this study, decision trees, neural networks and random forests were used in the model, and the company data from 2016 to 2019 reported by TEJ Economics were used to compare accuracy, type one error and type two error to find out the best financial early warning model.