基于随机森林算法的美国芝加哥空气质量预测研究

A Study on Air Quality Prediction in Chicago, USA Based on Random Forest Algorithm

摘要: 随着城市化的加速，空气污染问题日益严重，因此准确预测影响空气质量的污染物含量对于改善城市空气质量具有重要意义。本文基于随机森林算法建立了空气质量预测模型。该模型仅利用与空气质量指数(AQI)相关性较强的污染物历史数据，即可对未来的空气质量进行短期预测，模型操作简单，预测结果精确度也较高。首先，选取了美国芝加哥市2019年1月至2021年1月的空气质量综合指数，对该城市的AQI数据进行相关性分析，识别出与AQI相关的主要污染物指标。然后，采用随机森林算法建立空气质量预测模型。实验结果表明，本文建立的空气质量预测模型对于芝加哥城市O₃浓度的训练集上的R²达到了0.92，验证集上的拟合优度为0.87，对于PM_2.5、NO₂、SO₂等其他主要污染物也有较好的拟合效果。本研究基于AQI提取的相关性污染物指标的随机森林预测模型能显著提升空气质量预测的准确性，能够为空气质量管理提供有效的决策支持。

Abstract: With the acceleration of urbanization, air pollution has become increasingly severe. Therefore, accurately predicting pollutant concentrations that affect air quality is crucial for improving urban air quality. This paper establishes an air quality prediction model based on the Random Forest algorithm. This model can make short-term predictions of future air quality using only historical data of pollutants strongly correlated with the Air Quality Index (AQI). The model is simple to operate and delivers highly accurate prediction results. First, the comprehensive air quality index data for Chicago, USA, from January 2019 to January 2021 were selected. Correlation analysis of the city’s AQI data identified key pollutant indicators correlated with AQI. The Random Forest algorithm was then employed to build the air quality prediction model. Experimental results demonstrate that the proposed air quality prediction model achieves a coefficient of determination of 0.87 for O₃ concentrations in Chicago, with comparable fitting performance for other primary pollutants, including PM_2.5, NO₂, and SO₂. This random forest model, trained on correlation-based pollutant indicators extracted from AQI data, significantly enhances the accuracy of air quality forecasting and provides effective decision support for air quality management.