Predicting the Outcome of Construction Change Disputes Using Machine-Learning Algorithms

Abstract

Construction disputes are among the most stressful events that may occur throughout the course of a project. Construction executives are increasingly seeking new means to avoid and resolve disputes. Artificial intelligence may be utilized to predict court judgments by uncovering hidden links between interconnected dispute factors, giving disputing parties a better insight on their case position and likely possible outcome. This paper investigates the change order disputes by creating a list of legal factors on which the court rulings were based for previously similar cases in order to determine the likelihood of a potential outcome for a future claim. Various machine-learning models are utilized and tested to determine the best conforming algorithm. These models are evaluated using confusion matrix based on their accuracy, precision, recall, and sensitivity. This study found that the random forest algorithm rendered the best overall performance and achieved (95.0%) prediction accuracy. The model developed in this research may be utilized as a practical means by disputing parties to evaluate and decide whether to file a claim or to settle it privately to resolve the disputes more efficiently for construction dispute negotiation purposes.