Classification of Air Quality Network based on Meteorological and Pollutant Factors

Tüysüzoğlu G., Birant D., Kut A., Pala A.

2nd International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA), Turkey, 26 - 27 June 2020, pp.128-133 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • Doi Number: 10.1109/hora49412.2020.9152910
  • Country: Turkey
  • Page Numbers: pp.128-133
  • Keywords: air quality monitoring, classification, ensemble learning, machine learning, MONITORING STATIONS
  • Dokuz Eylül University Affiliated: Yes


In order to measure air pollution, to provide air quality control for the dangerous regions, and to provide stability in other regions, many developed countries have established air quality measurement stations. All of these sites have meta-data information containing the type of station such as urban, rural or industrial according to the location of the corresponding station or the characteristic specialty of the surrounding region. The classification of these stations under certain categories is an important process because if the type of station is known, the institutions that provide environmental auditing will transfer the resources appropriate to these regions and adequate control will be provided about the air quality for them. For this purpose, it was aimed in this study to determine to which class to be assigned when a new station is to be set up by taking into account the past pollutant concentrations and meteorological factors. In the experimental studies, different classification algorithms and their ensemble models are compared with our ensemble learning model "Enhanced Bagging (eBagging)" to classify 21 sites in the air quality monitoring network of Turkey. As a consequence, the eBagging ensemble learning algorithm combined with C4.5 significantly outperforms single classification models and their ensembles by better classifying the monitoring stations in terms of the air pollutant concentrations and meteorological data.