Comparative Analysis of Forecasting Construction Costs via Automated Machine Learning Approach and Inclusion of Economic Index

Construction cost prediction in the early phase of construction projects is crucial for setting the overall budget and facilitating bidding and contracting processes. While research efforts have been made by adopting machine learning (ML) algorithms to develop robust cost prediction models, they often heavily rely on professionals’ domain-specific knowledge and manual work (e.g., hyperparameter optimization). Also, despite the relationship between the economic index and construction costs, it has been minimally integrated into the dataset used to train the algorithms. A promising alternative is to leverage an automated machine learning approach (AutoML) that incorporates a market-sensitive indicator, the Construction Cost Index (CCI). To this end, this paper introduces an AutoML-centered construction cost forecasting approach and demonstrates its effectiveness through internal/external validation, the inclusion of CCI, feature importance analysis, and an investigation of the trade-off between the number of input features and computing efficiency.