Safety performance prediction and modification strategies for construction projects via machine learning techniques

Purpose: Unsafe behavior accounts for a major part of high accident rates in construction projects. The awareness of unsafe circumstances can help modify unsafe behaviors. To improve awareness in project teams, the present study proposes a framework for predicting safety performance before the implementation of projects. Design/methodology/approach: The machine learning approach was adopted in this work. The proposed framework consists of two major phases: (1) data collection and (2) model development. The first phase involved several steps, including the identification of safety performance criteria, using a questionnaire to collect data, and converting the data into useful information. The second phase, on the other hand, included the use of the decision tree algorithm coupled with the k-Nearest Neighbors algorithm as the predictive tool along with the proposing modification strategies. Findings: A total of nine safety performance criteria were identified. The results showed that safety employees, training, rule adherence and management commitment were key criteria for safety performance prediction. It was also found that the decision tree algorithm is capable of predicting safety performance. Originality/value: The main novelty of the present study is developing an integrated model to propose strategies for the safety enhancement of projects in the case of incorrect predictions.