Random Forest-Based Decision Tree Framework for Hazard Management in University Laboratories

Abstract

This study explores the use of Decision Tree (DT) algorithms for detecting potential hazards in laboratory operations by analyzing a synthetic dataset modeled on historical accident reports. The dataset simulates mishaps and near-miss incidents in university laboratories, incorporating detailed descriptions of behaviors and risks. Feature extraction techniques like Principal Component Analysis (PCA) are used to train and test DT models. The Random Forest-based Decision Tree (RF-based DT) model demonstrates superior performance compared to traditional methods. Implemented in Python, the model predicts chemical hazards with 92.3% accuracy, 93.1% precision, 93.5% F1 score, and 91.4% recall. These results confirm the model's reliability for risk detection and management in laboratory settings.