Machine Learning for Air Quality Classification in IoT-based Network with Low-cost Sensors

Abstract

Air pollution is a rising problem, with its effects being especially severe in urban and industrial areas. A constant and local monitoring of air quality, and the suitable presentation of the results to population, demands deployment of large-scale IoT-based monitoring networks in which low-cost, low-quality sensors would be predominantly used. However, the inherent measurement errors could incur large AQI (Air Quality Index) calculation error. Also, appropriate presentation of air pollution demands that measurements of air pollutants’ concentrations are classified into Air Quality Classes, thus making the classification task for AQI of large interest. In this paper we analyzed a wide variety of Machine Learning (ML) and Deep Learning (DL) models in order to solve classification task for AQI, but under the assumption of low-cost sensor deployment in the real-world application. The results of comprehensive analysis suggest that DL models designed, optimized and tested in this paper present a viable and the most suitable solution under these demands.