Prediction of Safety Performance by Using Machine Learning Algorithms: Evidence from Indian Construction Project Sites

Abstract

The construction industry in India happens to be the second most contributor to its gross domestic product (GDP) but high rates of accidents and fatalities have tarnished the image of the industry in India. To enhance the importance and alertness among the stakeholders in construction project sites, the present study proposes a framework for predicting safety performance. In this retrospective study, the data pertaining to the 69 construction project sites across India from January, 2021, to July, 2022 was analysed. The data analysis was conducted in two phases, in the first phase of the study the efficiency of project sites was computed by implementing data envelopment analysis (DEA). In the second phase, the results of the first phase are utilized to predict the safety performance of construction sites by applying four machine learning (ML) algorithms. In the first phase of the study, three input and three output variables were considered to compute the efficiency of the project sites. Results of four ML classifiers revealed that the random forest classifier with high recall percentage of 95.0 is considered the best in predicting the safety performance. Finally, the results indicate that the ML classifiers enable a good accuracy level in predicting the safety performance of project sites. Among the four ML classifiers, notably the Random Forest Classifier enables identifying the inefficient project sites and advising the site management to implement control measures. Finally, a safety performance prediction tool was developed to understand the results.