Time Series Analysis for Predicting PM2.5 Concentration in Shanghai based on Machine Learning

Xuede Huang; Lingling Liu; Zhonglin Lin; Lingling Chen

doi:10.54097/5mbbnd17

Authors

Xuede Huang
Lingling Liu
Zhonglin Lin
Lingling Chen

DOI:

https://doi.org/10.54097/5mbbnd17

Keywords:

Time Series Analysis, PM2.5 Prediction, Machine Learning, Model Comparison

Abstract

This study aims to use three different machine learning models -- ARIMA, LSTM and Random Forest -- to predict PM2.5 concentration in Shanghai. By analyzing and comparing the performance of these models in practical applications, we find that each model has its own unique advantages and limitations. The LSTM model performs best in processing complex multivariate time series data, showing high accuracy and excellent data processing capabilities, and its RMSE and MSE indicators are superior to other models. The random forest model excelled in feature importance analysis, providing valuable insights into understanding the drivers of changes in PM2.5 concentration.

Downloads

Download data is not yet available.

References

[1] Pan Zhengtong. Predictive Analysis of China's cargo traffic volume based on Machine Learning time series [J]. China Storage and Transportation,2022,(01):85-86.]

[2] Jiang Yuan, Yang Bo, Zhao Donglai, et al. Mobile network traffic prediction technology based on time series analysis and machine learning [J]. Internet of Things Technology, 2019, 10(06):42-45.

[3] FU Zhiou, Zhou Yang, Chen Cheng, et al. [3] Fu Z, Zhou Y, Chen C, et al. Application of time series analysis and machine learning method in predicting the incidence trend of pulmonary tuberculosis [J]. China Health Statistics, 2019,37(02):190-195.