基因演算法與類神經網路在財務危機預測上之應用

Application of Genetic Algorithms and Neural Network on Financial Crisis Prediction

鄭書甫(Shu-Fu Cheng)

指導教授 ： 姚舜暉 ； 陳寧馨

長榮大學/管理學院/經營管理研究所/碩士(2011年)

https://doi.org/10.6833/CJCU.2011.00020

摘要

隨著企業財務發生危機的頻率來越高，財務危機預警也越來越受到重視，而財務危機預警模型的研究早已行之有年，本研究首先將許多較常使用的財務預警模型進行整理，並利用Logit模型、類神經網路模型與基因演算法模型，分別進行財務危機預測，比較各模型預測的結果。結果顯示，以類神經網路模型的預測準確度最高，在全期間模型中預測正確率最高可達97.5%；其次為基因演算法，預測正確率可達81%； 而Logit模型的預測正確率則是低於兩者。
本研究進一步針對基因演算法模型，進行變數修正，並比較變數修正前、後的財務危機預測準確度。結果顯示，在財務危機前一年與前二年的期間，修正後的基因演算法模型準確度高於修正前，而在財務危機前三年與全期間則是低於修正前的基因演算法模型。修正後的基因演算法模型預測準確度仍然無法高於類神經網路模型的預測結果。
總結本研究結果，在探討的幾種模型中，以類神經網路模型在財務危機預測準確度最高。

關鍵字

基因演算法 ； 類神經網路 ； 財務危機預測

並列摘要

As the frequency of enterprise financial crisis gets higher, early
warnings of financial crisis are paid more and more attention. There have
been studies on financial crisis early-warning models for several years.
This study first collated and analyzed many commonly-used financial
early-warning models. Secondly, predictions of financial crisis were
carried out by means of three kinds of models, i.e. Logit model, artificial
neural network model and genetic algorithm model, and then the results
of the predictions were compared. The results showed that the accuracy
predicted by the artificial neural network model was the highest; in the
whole period of prediction, the accuracy rate was up to 97.5%. The
second high prediction accuracy was by genetic algorithm with an
accuracy rate 81%. The accuracy rate of the Logit model prediction was
lower than of the other two.

This study further performed variable correction on the genetic
algorithm model to compare the results of financial predictions before
and after correction. The results showed that before the financial crisis
during the previous two years before the financial crisis, the accuracy by
the revised genetic algorithm model was higher than before correction,
while the accuracy was lower than that by the unrevised genetic
algorithm model in the period of the time three years before the financial
prediction and in the whole period of time. The prediction accuracy by
the revised genetic algorithm model still could not get over the neural
network model.

In conclusion, of the various kinds of models disused here, the
artificial neural network model demonstrated the highest accuracy
prediction of financial crisis.

並列關鍵字

genetic algorithm ； artificial neural network ； financial crisis prediction

參考文獻

4. 李基賢(2010)。中小企業信用評等模型之評估—財務比率與專家經驗。國立中興大學碩士論文。

7. 林群凱(2005)。上市公司財務危機預警模式-以非財務資訊及不同預測模型建構。國立成功大學碩士論文。

1.Altman, E. I. (1968). “Financial Ratios, Discriminant Analysis and the

Prediction of Corporate Bankruptcy”, Journal of Finance,

23(4),pp.589-609.