Prediction of effect of wind speed on air pollution level using machine learning technique

Abstract Air pollution is one of the most challenging issues poses serious threat to human health and environment. The increasing influx of population in metropolitan cities has further worsened the situation. Quantifying the air pollution experimentally is quite a challenging task as it depends on many parameters viz., wind speed, wind temperature, relative humidity, temperature etc. It requires the investment of huge money and manpower for controlling air pollution. Machine learning technique-based computer modelling reduces both of the parameters. In the present work, the dependence of air pollution level on wind speed and temperature has been taken up using machine learning in the form of ANN and LSTM model. The recorded data of air pollution level (PM2.5) is collected from a measurement station of Lucknow city situated at Central School, CPCB. The data is used in an Artificial Neural based network and in an LSTM model to predict suitably the level of air pollution for a known value of average wind speed and temperature without experimental measurements. LSTM model is found to predict the pollution level better than ANN for the developed ANN networks.