Abstract—Particulate matter (PM) can harm human health
by causing lung cancer, pneumonia, or cardiovascular disease.
There is a growing awareness of dangerous PM among people
and governments. In order to prepare for the risk, the
prediction performance of PM is important. Therefore, many
kinds of research are developing various prediction models.
Among the models, LSTM based models show the best result
and it uses various auxiliary data, including spatial features to
improve performance. However, spatial features can be
depreciated because all input data has to be unfolded to 1D
vector. In this paper, we apply Convolutional LSTM to our
model to take advantage of the spatiotemporal relation of the
wind and PM forecasting problem. Also, we add CNN to extract
temporal features of the dataset on our model in parallel.
Finally, we combine both Convolutional LSTM and CNN to
predict more accurate PM concentration. In the experiment, we
compared this model with LSTM and CNN-LSTM models in
previous studies. At the result, the hybrid model showed the
best performance.
Index Terms—Deep learning, convolutional long short-term
memory (ConvLSTM), CNN, particulate matter prediction.
Seonggu Lee is with the Department of Human ICT Convergence,
Sungkyunkwan University, Suwon, Korea (e-mail: dltjdrn@skku.edu).
Jitae Shin is with the Department of Electrical and Computer Engineering,
Sungkyunkwan University, Suwon, Korea (e-mail: jtshin@skku.edu).
[PDF]
Cite:Seonggu Lee and Jitae Shin, "Hybrid Model of Convolutional LSTM and CNN to Predict Particulate Matter," International Journal of Information and Electronics Engineering vol. 9, no. 1, pp. 34-38, 2019.
