Hybrid Method Combining Variational Mode Decomposition and Deep Neural Networks for Predicting PM2.5 Concentration in China

This project presents a hybrid approach combining Variational Mode Decomposition (VMD) and Deep learning model for predicting PM2.5 concentrations in China. The dataset consists of 4,136 records containing air quality and weather data, including PM2.5, PM10, NOx, CO, Ozone, temperature, humidity, wind speed, and solar radiation. Several missing values, particularly for certain volatile compounds like Xylene and Eth-Benzene, were identified and addressed during preprocessing. The hybrid model integrates VMD to decompose complex time series data and employs advanced DNN architectures, such as LSTM, RNN, GRU, and linear regression, for accurate forecasting. The primary objective is to develop a reliable model for PM2.5 concentration prediction, which can be pivotal for understanding air pollution dynamics. The study demonstrates the potential of combining signal processing techniques with deep learning models to enhance predictive accuracy in environmental datasets.

Keywords: Variational Mode Decomposition, Deep Neural Networks, PM2.5, LSTM, RNN, GRU, air quality prediction, machine learning, environmental monitoring, data preprocessing.