營造公司違約機率預測—運用強化訓練之Logit模型

A logit model using enforced training procedure to predict default probability in the construction industry

梁世英(The-Anh Luong)

指導教授 ： 曾惠斌

國立臺灣大學/工學院/土木工程學研究所/碩士(2010年)

https://doi.org/10.6342/NTU.2010.03095

摘要

無資料

關鍵字

營造公司 ； 違約機率預測 ； 強化訓練 ； 回歸模型

並列摘要

Construction industry has been playing an important role in the growth of many countries. The financial health of construction companies is a leading concern of many stakeholders such as the government, project owners/contractors, financial lending institutions and other investors. Therefore, many previous studies based on financial ratio analysis have been conducted for construction firm default prediction.
However, most of these studies use the sample-matching method which produces sample-selection biases. To avoid this shortcoming, we utilize all available firm-years data of U.S. construction companies for our sample set, which includes 51 defaulted and 1,371 non-defaulted samples, to build our logit model. Yet this is referred to the matter of between-class imbalance. To solve the between-class imbalance when analyzing data, our model performs enforced training on the defaulted sample set by replicating these samples several times. After that, all financial variables relative to default prediction in the construction industry as well as significant variables selected by MDA stepwise and Pearson correlation methods are put into the model for comparison.
The experimental results of this study point out that the discriminating power of the logit model can be greatly improved within 12 times of enforced training (the area under ROC curve AUC=0.7996 compared to AUC=0.7279 without “enforced”). Moreover, the logit model with enforced training procedure is more convenient to use because it is relatively independent of the selection of variables. Therefore, we recommend the proposed enforced logit model as an alternative to the existing prediction models for construction firms.

並列關鍵字

default probability prediction ； construction industry ； financial ratio ； logit model ； between-class imbalance ； enforced training

參考文獻

[1] Abidali, A., & Harris, F. (1995). A methodology for predicting company failure in the construction industry. Construction Management and Economics , 13 (3), 189-196.

[2] Agarwal, V., & Taffler, R. (2008). Comapring the performance of market-based and accounting-based bankruptcy prediction models. Journal of Banking and Finance , 32, 1541-1551.

[3] Altman, E. (1968). Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. Journal of Finance , 23 (4), 589-609.

[4] Beaver, W. (1966). Financial ratios as predictors of bankruptcy. Journal of Accounting Research , 4, 71-102.

[5] Chang, F., Chang, I., & Wang, Y. (2007). Enforced self-organizing map neural networks for river flood forecasting. Hydrological Processes , 21, 741-749.