Intelligent Contractor Default Prediction Model for Surety Bonding in the Construction Industry

Construction is a risk-filled, uncertain, and dynamic environment. Contractor default is a critical risk that can influence the outcome of projects in the construction industry. Construction project owners and other stakeholders look for methods to predict the potential of contractors to default, in order to avoid awarding contracts to high-risk contractors. One of the most effective tools for project owners to mitigate the risk of contractor failure is to transfer the risk of project completion to a surety company. The surety company conducts a comprehensive prequalification (underwriting) process to assess the possibility of contractor default. The prequalification process is done to evaluate any contractor, project, and contractual risks that may affect the contractor’s performance. The prequalification process involves evaluating various qualitative and quantitative evaluation criteria, many of which contain uncertainty and require subjective judgment.

This thesis demonstrates how fuzzy logic and expert systems techniques are integrated to develop a model able to help surety professionals in contractor default prediction for a specific construction project for bonding purposes. Building the contractor default prediction model (CDPM) included identifying, classifying, and providing a comprehensive, detailed list of the evaluation criteria for contractor and project prequalification. Numerical scales were defined for the quantitative evaluation criteria, and rating scales, using reference variables, were developed to quantify the qualitative criteria. An important evaluation category, “contractor’s organizational practices,” was incorporated as input to the CDPM. The CDPM was built using the expertise of surety practitioners across Canada, and several different knowledge acquisition techniques were used. A novel methodology for finding a group consensus function that aggregates experts’ judgment scores to represent a common opinion was applied, in order to aggregate the experts’ inputs for the CDPM development. A methodology to apply two different optimization techniques, genetic algorithms and artificial neural network back-propagation, for the CDPM’s adaptation is presented. Finally, software for contractor default prediction, SuretyQualification, is developed.