Predicting Project Contingency in the Construction Industry Using Machine Learning Algorithms

Financing construction projects is lucrative but risky, with projects often running over budget due to unforeseen expenses. Lenders and owners require contingency funds as a safety net against such financial uncertainties, which are increasingly prevalent due to volatile material costs, extreme weather, and economic pressures in today’s construction market. Current methods for estimating contingency funds rely heavily on qualitative assessments and generally lack data-driven approaches. This research aims to fill this gap by employing machine learning to create a predictive model for estimating the range of contingency costs. Utilizing data from 150 construction projects, this study developed and compared the performance of four machine learning algorithms: k-Nearest Neighbors (KNN), Random Forest (RF), Artificial Neural Network (ANN), and Naïve Bayes (NB). The methodology includes data collection, preprocessing, and training the different algorithms. The results showed that the ANN is the most accurate by achieving an overall prediction accuracy of 93.33%. Such predictive capability is vital for assessing project risks and ensuring financial stability against unexpected costs. This study adds to the body of knowledge by proposing a quantitative, data-driven approach/tool to help decision-makers in construction finance in better estimating the ranges of contingency funds and ultimately contributing to better risk management in the construction sector.