A Multi-Layer Perceptron Approach For Estimating Daily Surface NO2 In Thiruvananthapuram City

Abstract

In this study, a machine-learning framework based on a multi-layer perceptron (MLP) algorithm is applied to estimate the daily values of air pollutant NO 2 in Thiruvananthapuram city of Kerala, India. The risk of human respiratory tract infections rises when exposed to high amounts of NO 2 [1]. Due to urbanization and its consequences, the air quality in the study region is getting deteriorated [2]. As a result, there is a pressing need for research and estimation of air pollutants like NO 2 in Thiruvananthapuram city. MLP is a supervised neural network model that is frequently used and it gains experience by learning to simulate the correlation between a set of input-output pairs [3]. This paper proposes a four-layer (i.e. one input, two hidden and one output) multi-layer perceptron neural network model for predicting the daily surface NO 2 values. Two year daily data (January 2018 to December 2019) is collected from Central Pollution Control Board, Government of India. The study utilizes 8 air pollutant parameters (PM 10 , PM 2.5 , SO 2 , NO, NO x , NH 3 , CO and Ozone) and 7 meteorological parameters (wind speed, wind direction, air temperature, solar radiance, relative humidity, atmospheric pressure and rainfall) in the model development. Due to instrumental errors, certain data are missing and such missing daily data records are excluded from