Construction disputes are among the most critical challenges in project management due to the complexity, uncertainty, and multi-stakeholder nature of construction projects, often leading to cost overruns, schedule delays, and contractual tensions. This study aims to develop a data-driven framework for proactive dispute management in construction projects using machine learning techniques. To achieve this, a set of variables related to project characteristics, contractual conditions, payment status, design changes, schedule performance, documentation quality, and managerial factors was considered as model input, while dispute occurrence and dispute severity were treated as output indicators. The proposed approach applies machine learning models to identify hidden patterns and nonlinear relationships among dispute-related factors and to improve the prediction of potential claim situations. The results indicate that delayed payments, design changes, ambiguity in contract documents, schedule slippage, and weak documentation are among the most influential factors contributing to dispute formation and escalation. The main novelty of this research lies in integrating predictive modeling with interpretable analysis to support a clearer understanding of risk drivers in construction disputes. The findings offer practical value for project managers, contractors, and employers by enabling earlier risk identification, better decision-making, and more effective preventive actions to reduce contractual conflicts and improve overall project performance.
Construction disputes are a persistent challenge in project-based industries due to the complexity of contractual relationships, frequent design changes, financial uncertainty, and the involvement of multiple stakeholders. Such disputes often lead to schedule overruns, increased costs, reduced productivity, and strained professional relationships among project participants. Despite the importance of dispute management, traditional approaches are often reactive and rely heavily on expert judgment or post-event analysis, which limits their effectiveness in early risk identification. In this context, the growing availability of project data and advances in machine learning provide new opportunities for developing proactive and data-driven dispute management systems. Therefore, this study aims to investigate the potential of machine learning techniques for predicting dispute occurrence and severity in construction projects and to identify the most influential factors contributing to dispute formation.
This research adopts a data-driven methodology based on machine learning modeling to analyze the risk of disputes in construction projects. A set of input variables was defined by considering project characteristics, contractual conditions, payment-related issues, design changes, documentation quality, schedule deviations, and managerial factors. These variables were used to predict dispute-related outputs such as the likelihood of dispute occurrence and its potential severity. The analytical process included data preparation, feature selection, model development, and performance evaluation using appropriate validation procedures. In addition, interpretability techniques were incorporated to better understand the influence of key variables on model outputs. This methodological framework makes it possible not only to improve predictive performance but also to provide transparent and practically meaningful results for decision-makers.
The findings demonstrate that machine learning models can effectively identify dispute-prone situations in construction projects by capturing complex and nonlinear relationships among contributing factors. Among the examined variables, delayed payments, design changes, ambiguity in contract documents, schedule slippage, and weak documentation emerged as the most influential predictors of dispute occurrence and escalation. These results are broadly consistent with previous studies that have highlighted the central role of financial, contractual, and managerial deficiencies in dispute development. However, the present study goes beyond descriptive assessment by quantifying the relative importance of these factors within a predictive framework. The results also indicate that interpretable machine learning can enhance managerial understanding of dispute dynamics and support more informed interventions before conflicts evolve into formal claims or legal disputes.
This study shows that construction disputes can be addressed more effectively through predictive and data-driven approaches rather than solely through conventional reactive practices. By integrating machine learning with interpretability analysis, the research provides a framework for early dispute risk detection and for identifying the main drivers behind dispute formation in construction projects. The main contribution of the study lies in moving from descriptive explanations of dispute causes toward an operational prediction tool with practical decision-support value. The results can assist project managers, contractors, and employers in improving documentation practices, reducing contractual ambiguity, managing changes more effectively, and taking preventive actions at earlier stages of the project lifecycle. Overall, the proposed approach contributes to proactive dispute management and supports better project performance in terms of time, cost, and stakeholder coordination.