Analysis of disputes arising from construction contracts: Prediction of dispute outcomes and costs using artificial intelligence-based models

Disputes are quite common in construction sector, which is regarded as a pivotal component of national economies in developing nations. Disputes in the sector can cause irreparable serious losses. Therefore, understanding and reducing disputes arising from construction contracts is crucial for minimizing losses and maintaining the economic stability provided by the sector. The main objective of the thesis is to identify the key factors in disputes by promoting amicable resolutions among the parties involved and develop tools that will mitigate losses. For this purpose, this study examined the causes, consequences, impacts, and dispute resolution methods utilized for contract-related disputes in the construction sector through a holistic approach. In this context, a questionnaire was developed, drawing upon a comprehensive literature review and expert opinions. The questionnaire data were examined by statistical analysis. Additionally, a model predicting the cost of disputes was developed using 181 data points obtained through the questionnaire, employing a hybrid approach based on machine learning and mutual information. Apart from these, within the scope of the thesis, disputes between the public and private sectors were specifically examined, and a model predicting the outcomes of these disputes was developed. Within this framework, with special permission from the High Science Board under the Ministry of Environment, Urbanization, and Climate Change, a model was created to predict the outcomes of disputes using backward elimination and machine learning techniques in the Python environment, based on 141 construction dispute decisions belonging to this board. As a result of the study, the analysis of the conducted questionnaire revealed that more disputes occur in construction works and lump-sum type of contracts, predominantly involving the contractor and the employer. It was also determined that contractor-related disputes are the most prevalent among general disputes, that negotiation is the most commonly employed method of dispute resolution, and that the majority of disputes are resolved within six months. The study presents the accuracy rates of the models created for predicting the outcomes and costs of disputes, and detects the key variables for these models. Furthermore, the variables essential to both prediction models were evaluated in comparison with existing literature studies. This study contributes to the literature by clarifying the significant current factors in disputes and their interrelations, shedding light on the outcomes of disputes and the prediction of their costs based on empirical data.