News

CMU Researchers Use Machine Learning to Fill Decade-Long PM2.5 Data Gaps

15 May 2026
Corporate Communication and Alumni Relations Center (CCARC)
A team of researchers from Chiang Mai University’s Department of Computer Science, Faculty of Science, has successfully developed a machine learning model capable of predicting retrospective PM2.5 concentrations over the past 10 years across eight northern provinces of Thailand, effectively addressing gaps caused by inconsistent historical monitoring data. This breakthrough is expected to significantly benefit epidemiological research on the long-term health impacts of air pollution, while also supporting public health initiatives and air quality management policy planning.
The PM2.5 crisis has caused severe long-term impacts on public health. However, research on its long-term health effects has been hindered by incomplete and fragmented PM2.5 monitoring data. To address this challenge, the research team employed machine learning techniques to predict retrospective PM2.5 concentrations from 2011 to 2020 using data collected from the Pollution Control Department’s air quality monitoring stations across Upper Northern Thailand.
The data used to develop the model included PM10 concentrations, pollutant gases such as CO2 and O3 (Ground-level Ozone), fire hotspot records, and meteorological factors including atmospheric pressure, precipitation, relative humidity, temperature, wind direction and wind speed. The team compared the performance of five machine learning algorithms: Multi-layer Perceptron Neural Networks (MLP), Support Vector Machine (SVM), Multiple Linear Regression (MLR), Decision Tree (DT), and Random Forests (RF).
The results revealed that the Random Forests (RF) algorithm demonstrated the highest predictive performance for PM2.5 concentrations, achieving a Root Mean Square Error (RMSE) of 6.82 ?g/m?, a Mean Percentage Error (MPE) of 4.33 ?g/m?, a Relative Percentage Error (RPE) of 22.50%, and an R? value as high as 0.93. These findings highlight the strong potential of machine learning in advancing environmental monitoring and public health research.
The study, titled ‘Long-term retrospective predicted concentration of PM2.5 in Upper Northern Thailand using machine learning models’, was conducted by Dr. Worawut Srisukkham from the Department of Computer Science in collaboration with Dr. Sawaeng Kawichai, Dr. Patumrat Sripan, Dr. Amaraporn Rerkasem and Professor Dr. Kittipan Rerkasem from the Research Institute for Health Sciences (RIHES), Chiang Mai University. The project aims to establish a comprehensive database on the long-term impacts of air pollution and contribute to sustainable solutions to the PM2.5 crisis in Northern Thailand.

Research and Innovation News Highlights COVID-19 and PM2.5

Gallery