Machine Learning Methods for Air Quality Monitoring

Abstract

Machine learning algorithms, and especially deep neural networks, provide universal estimator paradigms to approximate optimal solutions for arbitrary domain-specific problems. On the other hand, environmental-related problems that are a direct result of our rapidly changing climate are, nowadays, of the highest importance. Recently, the adoption of machine learning algorithms for environmental modeling has increased, especially in time series forecasting and computer vision. In this review, we attempt to provide a unified and systematic survey of the current machine learning algorithms used to solve multiple air quality monitoring tasks. We specifically focus on air quality modeling using satellite imagery and sensor device data. Lastly, we propose future directions with neural network modeling and representation learning.