Evaluating Contractor Financial Status Using a Hybrid Fuzzy Instance Based Classifier: Case Study in the Construction Industry

Abstract

Construction firms are vulnerable to bankruptcy due to the complex nature of the industry, high competitions, the high risk involved, and considerable economic fluctuations. Thus, evaluating financial status and predicting business failures of construction companies are crucial for owners, general contractors, investors, banks, insurance firms, and creditors. The prediction results can be used to select qualified contractors capable of accomplishing the projects. In this study, a hybrid fuzzy instance-based classifier for contractor default prediction (FICDP) is proposed. The new approach is constructed by incorporating the fuzzy K -nearest neighbor classifier (FKNC), the synthetic minority over-sampling technique (SMOTE), and the firefly algorithm (FA). In this hybrid paradigm, the FKNC is utilized to classify the contractors into two groups (“default” and “nondefault”) based on their past financial performances. Since the “nondefault” samples dominate the historical database, the SMOTE algorithm is employed to create synthetic samples of the minority class and therefore alleviates the between-class imbalance problem. Moreover, the FA is employed to determine an appropriate set of model parameters. Experimental results have shown that the proposed FICDP can outperform other benchmark methods.