|
Providing The Classification And Prediction of PM2.5 Pollutant Map Using Machine Learning Methods And Extracting Association Rules
|
Mohammad Reza Heydari   , Parham Pahlavani * , Behnaz Bigdeli |
| University of Tehran |
|
|
Abstract: (2638 Views) |
Air pollution is caused by the presence of various pollutants in the air, which is mostly related to the presence of particulate matters, especially particulate pollutant concentrations which are smaller than 2.5 microns (PM2.5). Predicting and identifying the infected areas will help us in managing and planning. Therefore, in order to identify these places, it is necessary to provide the maps of classification and the prediction maps of the PM2.5 pollution. The Supervised Machine Learning Methods used in this study, were Support Vector Machine, Multilayer Neural Network, and Decision Tree for classifying and predicting the PM2.5 pollutant maps in Tehran city. Moreover, to identify the effect of the spatial parameters, the Association Rules Mining Method was used. The Support Vector Machine Method with 87.3 percent for overall accuracy and 81.5 percent for the Kappa index was selected as the best classifier. This method was used to predict the concentration of the pollutants on the third day, which was able to predict the third day with 80.7 percent for the overall accuracy and 71.1 percent for the Kappa index. The findings indicate that the Support Machine Vector Method performs modeling and predicting with higher accuracy than the other methods. Attention to the influence of the spatial parameters in stronger association rules, the amount of the pollution of the nearest two neighborhoods, topography, temperature, air pressure, rainfall, intensity of air inversion, relative humidity, wind speed, wind direction, month of the year, day of the week, hour of the day had the greatest impact on determining the pollutant class.
|
|
| Keywords: Air Pollution, PM2.5 pollutant, Spatial parameters, Supervised Machine Learning, Associative Rules. |
|
|
Full-Text [PDF 1212 kb]
(254 Downloads)
|
Type of Study: Research |
Subject:
GIS
Received: 2022/05/22 | Accepted: 2023/05/14 | ePublished ahead of print: 2023/05/15 | Published: 2023/05/21
|
|
|
|
|
|
|
| Send email to the article author |
|
|
