碩士 / 實踐大學 / 企業管理學系碩士班 / 96 / Among the literatures concerning the construction of financial crisis alert model, most studies used all samples to construct the financial crisis differentiation model, and seldom divided the samples into in-sample and out-of-sample data for validation. Since out-of-sample forecast is the main purpose of research analysis, this study divided the samples into two groups, which are experimental group and validation group. It first used Pearson Product Moment correlation coefficient analysis to find out whether the variables have high correlation, and used factor analysis to extract representative latent factors. It then conducted discriminate analysis and logistic regression analysis to construct a financial crisis alert model for companies of shell reactivation.
The variables used in this study were mostly data obtained from the Market Observation Post System. Considering the bankrupted companies were mostly due to improper use of assets, this study adopted 25 financial ratios as research variable based on the cash flow indices suggested by Wang (1999) and Shih (2001). The results are as follows:
1.The first phase result of the alert model is based on the data of the experimental group: the assessed analytical error classification rate and hit rate are 3.2% and 96.8%, respectively, indicating that the assessment probability function has good predicative ability; the overall accuracy of the logistic regression analysis is 90.00%, indicating that the model has good predicative efficiency; the hit rate on normal companies is 93.33%, and that on those with financial crisis is 86.67%.
2.The second phase result is based on multiplication of the data samples of the validation group and the weight coefficient matrix parameter estimated by the experimental group. The obtained factor fractions were validated based on the model constructed in the first phase based on the data of the experimental group, in order to verify the accuracy of the alert model. The results are: the analysis error classification rate and hit rate are 3.60% and 96.40% in thinking about prior probability; the hit rate is as high as 96.40%, indicating that the model established by this study has high reliability; the overall accuracy of the logistic regression analysis is 80.00%, indicating that the model has good predicative efficiency; the predicative accuracy for normal companies is 60.00%, and that for companies in financial crisis is 100.00%.
3.Based on the results of the validation group, the predicative classification for normal companies match the actual classification; the predicative classification for financial crisis companies in the validation group of logistic regression analysis also matches actual classification; both methods have their advantages and weakness, and thus are complementary.
4.The results of the validation group showed that both discriminate analysis and logistic regression analysis have good predicative abilities (hit rate≧60%), of which, the discriminate analysis is better. The models constructed by those two methods could be used to alert companies of shell reactivation the signs of financial crisis, so that investors and supervisory bureaus can be watchful and adopt proper monitoring measures, or the information could be provided to financial institutions as references during loan review.
| Identifer | oai:union.ndltd.org:TW/096SCC00121029 |
| Date | January 2008 |
| Creators | Linda.liu, 劉淑梅 |
| Contributors | 方國榮 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 88 |
Page generated in 0.0125 seconds