Persistent, Bioaccumulative, and Toxic (PBT) chemicals, such as certain pesticides, industrial pollutants, and some plastic additives, pose significant environmental and human health risks. These substances are characterized by their long persistence in the environment, resisting natural degradation processes over extended periods. They bioaccumulate in the food chain, starting from small organisms like plankton and gradually concentrating in higher - level predators, including humans. For example, DDT, a well - known PBT chemical, was widely used in the past and is still detected in the environment decades after its ban. Due to their toxicity, PBT chemicals can cause a range of health issues, from disrupting the endocrine system to increasing the risk of cancer. In this study, to better identify such hazardous substances, machine learning techniques were harnessed to develop a Quantitative Structure–Activity Relationship (QSAR) model for predicting the PBT properties of chemicals. A dataset of PBT and non - PBT chemicals was assembled and preprocessed, with molecular descriptors derived from SMILES representations. One - hot encoding was utilized for classification labels, and eight machine learning models, including Random Forest, XGBoost, and Multi - Layer Perceptron, were trained for PBT classification. Model performance was evaluated using accuracy and area under the curve (AUC) metrics, with Random Forest demonstrating the best predictive ability. These results highlight the potential of machine learning in the high - throughput screening of hazardous chemicals, offering support for regulatory decision - making and environmental risk assessment.