Development of a Classification Framework for Construction Personnel’s Safety Behavior Based on Machine Learning

Different sets of drivers underlie different safety behaviors, and uncovering such complex patterns helps formulate targeted measures to cultivate behaviors. Machine learning can explore among behavioral data. This paper aims develop a classification framework for construction personnel’s behaviors with machine algorithms, including logistics regression (LR), support vector (SVM), random forest (RF), categorical boosting (CatBoost). The has three steps, i.e., data collection preprocessing, modeling algorithm implementation, optimal model acquisition. For illustrative purposes, five common sample Hong Kong-based personnel are used validate the framework. To achieve high performance, this employed combinative strategy, consisting feature selection, synthetic minority over-sampling technique (SMOTE), one-hot encoding, standard scaler classifiers classify multi-objective slime mould (MOSMA) optimize parameters in classifiers. Results suggest that strategy CatBoost–MOSMA achieves highest performance maximum average scores, area under curve receiver characteristic operator (AUC) ranging from 0.84 0.92, accuracy 0.80 0.86, F1-score 0.79 0.86. From model, unique set important features was identified each behavior, ten out 46 input indicators were found all Based on findings, study advocates using future behavior research makes concrete suggestions