Directorate of High Technics Board Decisions in Public Construction Projects: A Machine Learning Approach to Dispute Resolution

Disputes arising in public construction projects can lead to time and cost overruns, which negatively impact project processes. This study developed a machine-learning model to predict such disputes by analyzing decisions from the Directorate of High Technics Board of Public Works. Through a literature review and content analysis, a total of 18 legal factors were identified. Using the CorrelationAttributeEvaluation method, 11 out of the 17 input legal factors were selected as target legal factors for inclusion in the model. The MultiBoostAB algorithm demonstrated the best performance, achieving an accuracy rate of 70.41%. The content analysis method employed in the model ensured an objective identification of legal factors specific to the legal context and played a significant role in interpreting the decision texts. However, the limited size of the dataset was identified as a limitation, affecting the model’s generalizability. Future studies are recommended to expand the dataset, apply over-sampling techniques, and integrate natural language processing methods. In addition, analysing decision texts using methods such as Natural Language Processing (NLP), BERT and RoBERTa to identify legal factors and establish prediction models can strengthen the identification of factors specific to the legal context and the accuracy of predictions.