1. Title, Journal and Authors
Title: Forecasting Air Pollution Particulate Matter (PM2.5) Using Machine
Learning Regression Models
Journal: Procedia Computer Science 171 (2020) 2057–2066
Authors: Doreswamy, Harishkumar K S1, Yogesh KM, Ibrahim Gad
2. Summary
This paper tests PM2.5 concentrations in Taiwan, where the mortality rate from cardiovascular and respiratory diseases is high among the many air pollution cities, to find the most suitable of the various methods of predicting fine dust with modern approaches with information on Taiwan Air Quality Monitoring data sets through 76 different stations from 2012 to 2017.
The experimental method will be divided into four stages.
First, Taiwan Air Quality Monitoring data sets divided the data obtained through 'particular data' and 'chronological data'. The "particular data" includes topographic information, while the "chronological data" includes focus of pollution or daily information, weekly information, and meteorological information.
Second, because these data may be missing, they are filled with missing data using 'Fourier Arrangements' or 'Spline Multinomial'.
Third, various machine learning algorithms are carried out for prediction performance.
Fourth, select the best algorithm for prediction through cross-validation such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Square Error (MSE), and Cooperative of Detection (R2).
3. Originality and Creativity
In this paper, as the predictive model for PM2.5, the Decision Tree Relator and the Gradient Boosting Regulator showed the best performance in training, and in the actual test, the Gradient Booster was the best. I wonder why there was a difference between the training stage and the test stage.
4. Application to research
5. Contact
Ma, Kyoung Rim / M.S. program
Environmental Systems Engineering Lab.
School of Earth Sciences and Environmental Engineering
Gwangju Institute of Science and Technology
1 Oryong-dong Buk-gu Gwangju, 500-712, Korea
Phone: 010-5058-6832
E-mail: dkvmfhelxp34@gm.gist.ac.kr