Water pollution, intensified by rapid industrialization and urbanization in China, poses significant environmental challenges. This study aims to predict water pollution levels using data from Chinese publicly listed companies spanning from 2007 to 2022. Employing the Gradient Boosting Regression Tree model, we focus on key pollutants: Chemical Oxygen Demand, Ammonia Nitrogen, Total Nitrogen, and Total Phosphorus. The dataset includes environmental and air pollution-related variables, with missing values imputed for consistency. Results indicate that ‘Year’ is the most influential feature across all pollutants, while ‘Dust’ and ‘COD’ are the least influential. Notable interactions between ‘Year’ and features like ‘NOx’ and ‘SO2’ were observed. The model shows strong predictive performance for NH3_N and TN, moderate for TP, and poor for COD, indicating areas needing further optimization. This study offers valuable insights for environmental management and policy-making, emphasizing the importance of temporal trends and feature interactions in predicting water pollution levels.