基于机器学习的城市空气污染预测
Urban Air Pollution Prediction Based on Machine Learning

随着工业化和城市化的加速发展，空气污染已成为全球面临的重大环境问题之一。本研究旨在通过对2010至2014年间逐小时收集的城市空气污染和气象数据进行深入分析，探索不同预测模型在空气质量(PM2.5浓度)预测中的表现。通过使用多重插补法处理数据集中的缺失值，研究构建了长短期记忆网络(LSTM)、ARIMA (1, 0, 1)和ARIMA (2, 0, 2)模型，对PM2.5浓度进行预测。结果表明，LSTM模型在评估指标中均显著优于ARIMA模型，验证了LSTM在处理复杂时间序列数据中的高效性。ARIMA (2, 0, 2)虽然性能优于ARIMA (1, 0, 1)，但与LSTM相比仍有较大差距。这项研究不仅提升了对空气质量动态的预测准确性，还为城市环境监测和公共健康管理提供了有力的数据支撑，进一步强调了高级机器学习技术在环境科学中的应用潜力。
With the accelerated development of industrialization and urbanization, air pollution has become one of the major environmental issues faced globally. This study aims to explore the performance of different predictive models in forecasting air quality (PM2.5 concentration) by conducting an in-depth analysis of hourly collected urban air pollution and meteorological data from 2010 to 2014. By employing multiple imputation methods to handle missing values in the dataset, this research constructs Long Short-Term Memory (LSTM) networks, ARIMA (1, 0, 1), and ARIMA (2, 0, 2) models for predicting PM2.5 concentrations. The results indicate that the LSTM model significantly outperforms the ARIMA models across all evaluation metrics, verifying the efficiency of LSTM in handling complex time series data. Although ARIMA (2, 0, 2) performs better than ARIMA (1, 0, 1), it still shows a considerable gap compared to LSTM. This study not only improves the accuracy of dynamic predictions of air quality but also provides robust data support for urban environmental monitoring and public health management, further highlighting the application potential of advanced machine learning technologies in environmental science.