Cluster-based prediction of air quality index

Abstract

The present world is facing the crucial issue of air pollution, which is threatening the human race and the environment equally. This situation requires effective monitoring of air quality and recording the pollution levels of different pollutants SO2, CO, NO2, O3 and particulate matters (PM2.5 and PM10). To achieve this, we need efficient prediction and forecasting models which not only monitors the quality of the air we breathe in but also forecast the future to plan accordingly. In this direction, various data mining methods are adopted for analysing and visualizing concentration levels of air pollutants using Data Mining on big data and data visualization. In this work, we have explored the partition-based clustering technique to extract the patterns from the air quality data. We have compared prediction methods, Auto-Regressive Integrated Moving Average (ARIMA) and Long Short-Term Memory (LSTM) on representatives of each cluster and also done forecasting for the year 2018. The results have proven that ARIMA works better than LSTM.