1. Title, Journal and Authors

 

Title: PM10 Density Forecast Model Using Long Short Term Memory

Journal: Institute of Electrical and Electronics Engineers

Authors: Jung-Hwan Park, Seong-Joon Yoo*, Kyung-Joong Kim, Yeong-Hyeon Gu, Keon-Hoon Lee, U-Hyon Son

 

2. Summary

 

This study used PM10 concentration data observed in Seoul from January 2005 to March 2016 and tried to predict future PM10 concentration using LSTM.

Traditional RNNs had less impact on the results of early data as data grew due to long-term dependency problems.

PM10 concentration data were preprocessed into 30days sequence data using the moving average technique. LSTM learned pre-processed data based on a sliding window to predict PM10 concentration.

ReLU was used as an activation function used for output, and the performance of the abnormal time series model was optimized using RMSProp. The mean square root error (RMSE) was used to measure the performance of the LSTM prediction model, resulting in a performance of 500% higher than linear regression and 100% higher than RNN.

It is expected that PM2.5 can be predicted if various factors are added.

 

Given that the performance of predicting PM10 concentrations is better than traditional methods, further studies of LSTM prediction models with additional factors such as climate and disease are expected.

 

3. Originality and Creativity

 

 

4. Application to research

 

 

5. Contact

Yeong Gyu Gu / Intern student

Environmental Systems Engineering Lab.

School of Earth Sciences and Environmental EngineeringGwangju Institute of Science and Technology

1 Oryong-dong Buk-gu Gwangju, 500-712, Korea

Phone : +82-10-6589-6653

E-mail : kududrb1@naver.com