| 摘 要: 空气污 染 问 题 日 益 突 出,已 经 成 为 阻 碍 健 康 城 市 建 设 的 关 键 问 题。为 解 决 这 一 问 题,采 用 基 于Transformer架构的空气污染预测模型 AirFormer,通过采集气象数据、空气污染物数据,并在此基础上融合交通流量和周边城建等信息,综合考虑时空动态变化因素,以提高预测的准确性。同时引入可解释性机制,通过特征权重进一步分析不同因素对空气污染的影响变化规律。实验结果表明,模型的均方根误差(RMSE)和平均绝对误差(MAE)分别为3.85和3.29,相较于其他模型降低9%~16%,在预测精度上有显著提升,且发现在早晚高峰时段交通因素对空气污染的影响突出。 |
| 关键词: 空气污染预测 Transformer架构 健康城市建设 时空相关性 时空数据 |
|
中图分类号: TP391
文献标识码: A
|
| 基金项目: 国家社会科学基金青年项目(25CGL002) |
|
| Air Pollution Prediction Integrating Trafficand Urban Construction Factorsfrom a Dynamic Spatio-Temporal Perspective |
|
ENG Shiyu1, YIN Pei1,2, YUAN Yixin1, QUAN Guanting1
|
(1.Business School, University of Shanghai for Science and Technology, Shanghai 200093, China;
2.School of Intelligent Emergency Management, University of Shanghai for Science and Technology, Shanghai 200093, China)
dengshiyu982460692@163.com; pyin@usst.edu.cn;; yyxyuanyixin@qq.com; 1163874272@qq.com
|
| Abstract: Air pollution has become increasingly prominent and a key issue hindering the construction of healthy cities. To address this problem, the AirFormer model based on the Transformer architecture is adopted for air pollution prediction. By collecting meteorological data and air pollutant data, and further integrating information such as traffic flow and surrounding urban construction, the model comprehensively considers dynamic spatio-temporal factors to improve prediction accuracy. Additionally, an interpretability mechanism is introduced to analyze the varying influence of different factors on air pollution through feature weights. Experimental results show that the model achieves a Root Mean Square Error (RMSE) and a Mean Absolute Error (MAE) of 3.85 and 3.29, respectively, representing a reduction of 9%-16% compared with other models, significantly improving prediction accuracy. Moreover, traffic factors are found to have a prominent impact on air pollution during morning and evening peak hours. |
| Keywords: air pollution prediction Transformer architecture healthy city construction spatio-temporal correlation spatio-temporal data |
