Risk-based completion cost prediction approach in construction projects utilizing machine learning

The construction industry is among the sectors exposed to frequent budget overruns. Therefore, accurately predicting costs to complete the construction projects is a vital point. Several research studies focus on cost estimation, construction risk factors, and their cost impact. Although they produced valuable prediction models for the completion cost of the projects, most of them mainly concentrated on the early stages of the construction. Limited studies produced approaches for completion cost estimation in the execution phase of the projects. Nevertheless, they do not implement total risk score effects in their models. Additional research is necessary to investigate risk-based completion cost prediction throughout the execution phase of construction. The main objective of this study is to provide an approach for the total risk score based completion cost prediction by using machine learning techniques without imposing excessive work. The proposed approach can be utilized at any point during the execution phase of a project to assess the impact of changes in total risk scores on completion costs. Furthermore, predicting the total completion cost using the total risk score simplifies the calculation and procedure rather than depending on breakdowns. To achieve this objective, a machine learning prediction approach was proposed to predict total completion cost based on total risk scores in construction projects. The proposed approach is applied to real-world cases to evaluate the accuracy of completion cost prediction based on risk scores using data from an international construction company. A total of 119 risk and cost data points from 11 projects were analyzed. Six prediction algorithms were employed, utilizing machine learning. Based on the outputs, it was determined that polynomial regression produced the most accurate predictions for available data. This research contributes to enhancing construction organizations' knowledge and planning capacities by quickly predicting project completion costs based on dynamic total risk scores at any time throughout the execution phase of the project.