Estimating Mobilization Costs in Infrastructure Projects Using Regressive Machine Learning Algorithms

Efficient project execution in the construction industry hinges on meticulous planning and resource allocation, with mobilization serving as a critical preparatory phase. This paper proposes a machine learning framework to predict a bidding range of mobilization costs in construction projects. Various regression models - including linear models, tree-based models, support vector machines, k-nearest neighbors, Bayesian models, neural networks, as well as other regressive models - were trained, tuned, and tested to predict the mobilization costs. This paper draws upon valuable data obtained from the Ohio Department of Transportation. Results indicate that Orthogonal Matching Pursuit and Ridge Regression are the most effective models for predicting minimum and maximum mobilization bid values, respectively. These findings enable contractors to propose realistic mobilization values, enhance early payment likelihood, and avoid negative cash flow during projects. The proposed model advances understanding of mobilization cost structures, fostering better planning, resource allocation, and risk minimization, thereby improving project efficiency.