An approach to forecast pollutants concentration with varied dispersion.

Globally, the rising level of air pollution is becoming a cause of serious concern. The situation has reached such an alarming situation that knowing the extent of air pollution well in advance has become an absolute necessity before we step out of our home. An advance prediction can help the urban travellers to know the possibility of enhanced pollution ahead of time at strategic locations of a city and thereby be useful in planning a less polluted route. Different cities of the world have pollutants levels with varied dispersion pattern. As a result, a generic prediction model is needed that can cater to all types of pollutants and which can offer better forecasts of pollution data irrespective of their dispersion levels. In this paper, the authors modelled recurrent neural network (RNN)-based bidirectional long short-term memory (Bi-LSTM) that forecasts pollutants concentration with least RMSE. The model is trained and tested on the pollution data of Delhi, the most polluted capital city in the world for the second consecutive year in 2019. Air pollutants like PM₁₀, PM_2.5, NO₂, O_3, and CO are considered that have a varied dispersion pattern. To test the efficacy of our predictions, the model is tested on real data obtained from the Central Pollution Control Board (CPCB) of India. The performance metrics are also generated to evaluate the performance of the proposed model. [ABSTRACT FROM AUTHOR]