臺灣博碩士論文加值系統

　　本研究以上市未上市及公司二組樣本，建立財務危機預警模型，探討下列主題：
　　1.危機公司與正常公司的財務狀況有何差異？
　　2.比較傳統計量模型（多變量模型或logit模型）與類神經網路模型的預測能力。
　　3.比較不同預測變數（公司財務比率、相對財務比率、公可財務比率變動、相對財務比率變動）的優劣。
　　所獲致結論如下：
　　1.正常公司的流動性、經營效率 、資本結構與獲利能力皆比危機公司佳。
　　2.類神經網路模型的前期樣本預測能力及型I誤差皆比logit模型好。
　　3.產業相對財務比率為較佳的預測變數，產業相對財務比率變動較差，且採用財務比率變動之觀念後，其預測效果不如原始財務比率。

　　The paper uses two different kinds of samples-- listing and non-listing companies --to establish predictiong model of corporate bankruptcy and discuss the following topics:
　　1.Are successful companies and bankrupt companies different in financial conditions?
　　2.which prediction model is better--traditional statistical model ( MDA or logit ) or neural network model?
　　3.which prediction variable is the best--corporate ratio( CR ), industry - relative ratio ( RR ), change of corporate ratio ( CR), and change of industry-relative ratio( RR )?
　　The results are as follows:
　　1.The financial conditions between successful and bankrupt companies are different, and performance of successful companies is better.
　　2.Neural network model is better than logit model, regarding total correct rate and type I error.
　　3.RR is the best prediction variable and RR is the worst. In additions, the perdition ability of CR and RR is worse than CR and RR.