Table 12 R2 and IA of the prediction approach in trend prediction.

From: A hybrid deep learning air pollution prediction approach based on neighborhood selection and spatio-temporal attention

Approach

R2

IA

1–12 h

13–24 h

25–36 h

37–48 h

1–12 h

13–24 h

25–36 h

37–48 h

CNN

0.632

0.520

0.430

0.302

0.918

0.900

0.883

0.845

LSTM

0.661

0.537

0.464

0.417

0.929

0.907

0.893

0.880

CNN-LSTM

0.783

0.715

0.709

0.636

0.952

0.941

0.937

0.926

ConvLSTM

0.795

0.721

0.698

0.646

0.954

0.940

0.936

0.923

STARes-SaLSTM

0.901

0.885

0.783

0.716

0.968

0.961

0.954

0.952

KSC-ConvLSTM

0.928

0.935

0.928

0.924

0.982

0.984

0.982

0.982