Uncovering the Causes of City Smog: A Machine Learning Approach

Air pollution, specifically fine particulate matter (PM2. 5), is a big problem in many Chinese cities. To tackle this, researchers used a machine-learning model that combines CatBoost and Tree-Structured Parzen Estimator (TPE) to study PM2. 5 levels in 297 cities from 2000 to 2021. They found that socioeconomic factors and industrial activities are the main culprits. The model was very accurate, with a score of 96. 44% in predicting PM2. 5 concentrations.

SHapley Additive exPlanations (SHAP) helped identify key factors. In 2000, high nitrogen oxide emissions, poor technology, and overcrowding were major issues in four heavily polluted cities. Over time, these cities saw improvements due to reduced nitrogen oxide emissions. Future strategies should focus on controlling population density and slowing mining development. This study offers a powerful tool to tailor strategies for each city.

Actions