| Research and application of the fields used by data mining techniques become more and more. This technique brings an outstanding economic benefit .Like quality control in manufacturing industry, prediction of the efficiency on operation, medication, diagnosis at hospital, customer relation management in service industry, especially credit risk control and financial crisis prediction in financial circles.With the rapid development of the market economy, it comes to be widespread that financial distress occurs recently. Constructing a complete crisis prediction system could find a crisis signal in time and take some efficient measures to improve condition of companies and avoid crisis occurring. It is important for investors, loaners, government department, banks and companies themselves. It could not only help good development of companies, but bring indispensable function to financial order and social stability.The author generalizes and analyzes the useful factors, in addition to financial variables, adds other important non- financial ones—corporate governance, market information and intelligence capital, to get more perfect results.The author collects 714 normal and 15 crisis firms of Taiwan in 2008. Using over-sampling technique to choose 1: 2, 1: 3, 1: 4 and using logistic regression analysis to construct prediction models of t-1 year, t-2 year, t-3 year. Then, proceeding simulation many times to compare their performances. Finally, taking classified matrix, receiver operating characteristic, K-S test and bootstrapping method to evaluate the power of forecast ability and stability and to provide choosing standard of models.This research finds every aspect has influence on financial distress. The prediction accuracy and stability of models are normally high. It shows that the prediction advanced by 3 years has a good result. We believe it could give more time for enterprises and investors to prepare and lower the rate of crisis occurring and reduce the damage when it occurs. |
PDF Full Text Request