Integration of Merton Model and Logit Model with Over-Sampling Technique to Predict Construction Contractor Default
Date Issued
2012
Date
2012
Author(s)
Nguyen, Tuan-Anh
Abstract
Since a construction firm goes bankrupt, not only the company but also other stakeholders will bear the great loss. However, past researchers normally excluded construction industry in their researches. Therefore, this study aimed to predict failure of construction contractor using an accounting-based model (logit model), a market-based model (Merton model), and a hybrid model.
Accounting information, while subject to human manipulations, can reflect some information not shown in stock price. In addition, option-based models are based on the presumption that all information is reflected on the company’s stock price, yet it is not always true in real life. To solve these disadvantages of above models, a hybrid default prediction model was developed.
In order to avoid bias due to the sample-matching method, all available data of firm-years are used to construct the default prediction models. Furthermore, replication and Synthetic Minority Over-sampling Technique (SMOTE), two over-sampling techniques, are proposed to tackle the imbalance problem in data set.
The empirical results show that over-sampling techniques could improve the predicting performance of not only logistic regression model but also hybrid model. Hybrid model using over-sampling techniques outperforms predicting ability than either Merton model or logit model. Additionally, in these two kinds of over-sampling techniques, SMOTE outperforms the prediction ability compare to Replication.
Subjects
construction contractor
default prediction
Logit model
Merton model
hybrid model
over-sampling
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-101-R99521254-1.pdf
Size
23.32 KB
Format
Adobe PDF
Checksum
(MD5):1d9387097c29f7e567c46ecdf7f2f257