Data Mining Paradigm in the Study of Air Quality

Abstract

Air pollution is a serious global problem that threatens human life and health, as well as the environment. The most important aspect of a successful air quality management strategy is the measurement analysis, air quality forecasting, and reporting system. A complete insight, an accurate prediction, and a rapid response may provide valuable information for society’s decision-making. The data mining paradigm can assist in the study of air quality by providing a structured work methodology that simplifies data analysis. This study presents a systematic review of the literature from 2014 to 2018 on the use of data mining in the analysis of air pollutant measurements. For this review, a data mining approach to air quality analysis was proposed that was consistent with the 748 articles consulted. The most frequent sources of data have been the measurements of monitoring networks, and other technologies such as remote sensing, low-cost sensors, and social networks which are gaining importance in recent years. Among the topics studied in the literature were the redundancy of the information collected in the monitoring networks, the forecasting of pollutant levels or days of excessive regulation, and the identification of meteorological or land use parameters that have the most substantial impact on air quality. As methods to visualise and present the results, we recovered graphic design, air quality index development, heat mapping, and geographic information systems. We hope that this study will provide anchoring of theoretical-practical development in the field and that it will provide inputs for air quality planning and management.