Machine learning methods to predict particulate matter PM _2.5

Area of Science:

• Environmental Science
• Computer Science
• Data Science

Background:

• Urban air pollution, particularly fine particulate matter (PM2.5), poses significant health risks globally, exacerbating conditions like asthma and cardiovascular disease.
• In Malaysia, research on machine learning for PM2.5 prediction is less developed compared to other air pollutants.
• Accurate air quality monitoring is crucial for public health and environmental management.

Purpose of the Study:

• To address the research gap in PM2.5 prediction in Malaysian smart cities using machine learning.
• To compare the effectiveness of different supervised machine learning techniques for forecasting PM2.5 concentrations.
• To identify the optimal model for accurate PM2.5 prediction to mitigate adverse health effects.

Main Methods:

• Utilized Malaysian air quality datasets from 2017-2018 for PM2.5 forecasting.
• Applied data preprocessing techniques including cleaning and normalization.
• Extracted informative features focusing on location and time, then trained Random Forest (RF), Artificial Neural Network (ANN), and Long Short-Term Memory (LSTM) models.

Main Results:

• The Random Forest (RF) model demonstrated superior performance, achieving an accuracy of 97.7%.
• Artificial Neural Network (ANN) and Long Short-Term Memory (LSTM) models achieved lower accuracies of 61.14% and 61.77%, respectively.
• RF's effectiveness was attributed to its ability to learn from random samples using decision trees with maximum voting for predictions.

Conclusions:

• Random Forest (RF) is the most effective supervised machine learning model for predicting PM2.5 concentrations in Malaysian smart cities among the tested algorithms.
• The study successfully identified a high-accuracy model for PM2.5 forecasting, contributing to better air quality management.
• Further research could explore additional features or advanced ML techniques for even more precise air pollution prediction.