Forecasting hourly PM2.5 based on deep temporal convolutional neural network and decomposition method

Fuxin Jiang, Chengyuan Zhang, Shaolong Sun, Jingyun Sun

Abstract: Highlights•A novel hybrid framework is built to forecast hourly PM2.5 concentration.•Steps: data collection, processing, decomposition, prediction, results analysis.•CEEMDAN is used for dealing the different timescales series of original PM2.5 data.•DeepTCN is employed for capturing information of multiple exogenous variables.•Empirical results verify its superiority over all the corresponding benchmarks.