A Comparative Study on Machine Learning Algorithms for Predicting Construction Projects Delay

Abstract

The perpetual occurrence of a global phenomenon – delay in construction industry despite considerable mitigation efforts remains a huge concern to its policy makers. Interestingly, this industry which produces massive amount of data from IoT sensors, building information modelling etc., on most of its projects daily is slow in taking the advantage of contemporary analysis method like machine learning (ML) which best explains factors that can affect a phenomenon like delay based on its predictive capabilities haven been widely adopted across other sectors. In this study therefore, a premise to compare the performance of machine learning algorithms for predicting delay of construction projects was proposed. To begin, a study of the existing body of knowledge on the factors that influence construction project delays was utilised to survey experts in order to obtain quantitative data. The generated dataset was used to train twenty-seven machine learning algorithms in order to develop predictive models. Results from the algorithm evaluation metrics: accuracy, balanced accuracy, Receiver Operating Characteristic Curve (ROC AUC), and f1-score indeed proved Perceptron model as the top performant model having achieved an accuracy, balanced accuracy, ROC AUC, and f1-score of 85%, 85%, 0.85 and 085 respectively higher than the rest of the models and unachieved in any previous study in predicting construction projects delay. Ultimately, this model can subsequently be integrated into construction information system to promote evidence based decision-making, thereby enabling constructive project risk management initiatives in the industry.