A Decision Tree Algorithm in University Laboratory Hazardous Chemicals Management and Alternative Technology Research

Abstract

Laboratory-based research activities include inherent hazards associated with managing hazardous chemicals and processes, requiring comprehensive risk assessment protocols to prevent accidents and injuries. In this study, we investigate the effectiveness of decision tree (DT) algorithms in detecting possible hazards in laboratory work using previous accident documents. The dataset consists of accidents and near-miss incidents reported from various university laboratory operations. Each report describes the nature of laboratory operations and related hazards. In our proposed model, we employ a Word2Vec-based DT algorithm, where Word2Vec transforms words into semantic vectors. The DT algorithm then utilizes these vectors to estimate hazard probabilities by recursively partitioning the data according to their characteristics for predicting the risks associated with university laboratory work. The proposed detection model has been implemented in a Python program. In the results assessment phase, we evaluate our proposed model's effectiveness in forecasting various chemical hazardous situations using numerous evaluation metrics such as recall, f1 score, precision and accuracy. We also carried out a comparison analysis with other traditional approaches. Our experimental findings demonstrate the reliability of the recommended framework.