Early Prediction of Construction Disputes: Decision Support Systems with Machine Learning Techniques

Abstract

This study aims to predict the outcomes of construction disputes before they proceed to litigation and to foster a constructive environment between parties. Within the scope of the study, a total of 24 legal factors; 14 legal factors were identified through extensive literature review and 10 legal factors were identified through content analysis. These legal factors were used in three stages: Pre-Litigation (A, B) and Post-Litigation. Legal factors with significant relationships were tested with 24 different machine learning algorithms. NB Tree, Logit Boost and LMT algorithms achieved 63.79%, 63.66% and 86.90% accuracy for models A, B and C, respectively.