Predicting financial distress of contractors in the construction industry using ensemble learning

•This study proposes financial distress prediction models based on ensemble learning.•The models provide two- and three-year ahead prediction of financial distress.•Performance was evaluated for contractors in South Korea from 2007 to 2012.•The models contribute to provide a financial early warning in the construction industry.•This model can help stakeholders to avoid damage due to financial crisis during a project. In the bid process, predicting whether the contractor will suffer a financial crisis during the construction project is vital to project owners and other stakeholders for identifying problems and taking strategic action. In this context, the models for predicting financial crisis of contractor have been extensively studied. However, the previous studies have been focused on predicting a financial crisis for one-quarter or one-year ahead of prediction point, even though the duration of projects are relatively long in the construction industry, usually exceeding one year. Moreover, despite the possibility of knowing the signs of financial crisis of a contractor through predicting financial distress, no attempt has been made to predict financial distress that contractor can suffer before reaching a financial crisis including highly visible legal events, such as bankruptcy, default, and delisting. This means that there is significant gap between those models and practical application in terms of the prediction period and definition of the financial crisis. This study proposes voting-based ensemble models that predict financial distress of contractor for two- and three-year ahead of prediction point using a finance-based definition of financial distress. The prediction performance of proposed model was evaluated using financial statements of contractors in South Korea from 2007 to 2012. The proposed models showed area under the receiver operating characteristic curve (AUC) values of 0.940 and 0.910 for predicting financial distress for each of the prediction years. By predicting financial distress of the contractor from the early stages of a construction project to the end stage with high accuracy, this model can help project owners and broad stakeholders to avoid damage due to financial crisis during a project.