Development of Nitrogen-Dioxide Concentrations Forecast System by Supervised Learning Approach in Kota City

Abstract

Many studies have been conducted in a row in current scenario that are concerned about the prediction of pollutants using supervised learning methods to check the pollution level as well as to control the trauma. As per the studies they are outperforming in comparison to any other approaches. In this paper a predictor model is developed for the prediction of prime pollutant from vehicle i.e. concentration of Nitrogen dioxide gas by taking the known set of input and target data that evolves predictions for the response of input and output data. Data of 2012,2013 and 2014 of a specific site of Kota city is taken into account to train, test and validate four different topologies of the neural networks namely Feed Forward Neural Network (FFNN), Layer Recurrent Neural Network (LRNN),Nonlinear autoregressive Exogenous (NARX) and Radial Basis Function Neural Network (RBFN ). Vehicles are the prime source of emission of this gas. A meaningful comparison between these topologies revealed that RBFN is a suitable topology for prediction engine.