Machine Learning-Based Temporary Traffic Control Cost Analysis

In a design-bid-build infrastructural project, the agency may use a lump-sum, or unit-price for temporary traffic control (TTC) items, while their cost is hard to estimate. This paper presents the research results of developing a machine learning model of the relationship between the TTC items’ cost with the project total cost and non-TTC items in infrastructural projects. In detail, 163 infrastructural projects’ data were collected for analyzing two research questions: first, the relationship between the TTC items with the project total cost and non-TTC items; second, the relationship between the TTC items’ payment option with the project total cost and non-TTC items. The results showed that the proposed feed-forward neural network model outperforms regression methods on classification tasks. It has a 36% accuracy in determining the TTC items’ cost as a percentage range of project total cost. Additionally, the proposed model has 94% accuracy in determining the TTC items’ payment options, when the information of the project total cost and the major non-TTC items’ information are known. With this research, the TTC items’ payment option for a new infrastructural project could be confidently decided, and the TTC items’ cost could be easily estimated as percentage ranges of the project total cost, which helps project owners and agencies to evaluate the quality of contractors’ bids.