Publication Date
12-2023
Advisor(s) - Committee Chair
Zhonghang Xia, Qi Li, Uta Ziegler
Degree Program
Department of Computer Science
Degree Type
Master of Science
Abstract
The Jing-Jin-Ji region of China is a highly industrialized and populated area of the country. Its periodic high pollution and smog includes particles smaller than 2.5 μm, known as PM2.5, linked to many respiratory and cardiovascular illnesses. PM2.5 concentration around Jing-Jin-Ji has exceeded China’s urban air quality safety threshold for over 20% of all days in 2017 through 2020.
The quantity of ground weather stations that measure the concentrations of these pollutants, and their valuable data, is unfortunately small. By employing many machine learning strategies, many researchers have focused on interpolating finer spatial grids of PM2.5, or hindcasting PM2.5. However, focusing on predicting concentrations will help reduce illness cases due to PM2.5.
This study proposes PM2.5 pollution forecasting using a LSTM-based neural network named Neighbor-Embedded Pollution LSTM, or “NEP-LSTM”, with two variants. The focus of this study is the questions: 1.) Does the proposed model NEP-LSTM, and its variant, outperform Random Forest trees when forecasting PM2.5 concentrations? And 2.) Does augmented neighboring weather stations’ data improve PM2.5 forecasting of these models?
This study concludes NEP-LSTM performs much better than Random Forests at forecasting, with a prediction error of only 6.06 μg/m3 PM2.5 RMSE, 75% less than Random Forests. The second variant of NEP-LSTM captures data better, due to containing an extra FNN block dedicated to processing neighboring weather stations. This difference yields an improvement in data representation 𝑅2 from 0.828 to 0.862. Feature importance analysis reveals insight into the worth of neighbor station data. Ultimately, NEP-LSTM variant 2 is chosen as the superior variant for forecasting PM2.5 concentrations for weather stations and utilizing augmented neighbor data.
Disciplines
Computer and Systems Architecture | Computer Engineering | Computer Sciences | Engineering | Environmental Monitoring | Environmental Sciences | Numerical Analysis and Scientific Computing | Physical Sciences and Mathematics | Systems Architecture
Recommended Citation
Pike, Richard, "Pollutant Forecasting Using Neural Network-Based Temporal Models" (2023). Masters Theses & Specialist Projects. Paper 3697.
https://digitalcommons.wku.edu/theses/3697
Included in
Computer and Systems Architecture Commons, Environmental Monitoring Commons, Numerical Analysis and Scientific Computing Commons, Systems Architecture Commons