PREDICTING PM2.5 CONCERTRATION USING ANN AND ASSESSING ITS HEALTH EFFECTS IN DELHI, INDIA

Abstract

The primary goal of the current study was to use an artificial neural network (ANN) to predict the linked health endpoint of PM2.5. The study area having 37 monitoring stations was taken and the air monitoring data for 2015 to 2019 was considered. The neural network utilised in this study has an output layer, an input layer with 8 parameters, and a hidden layer of neurons. At first the ANN was trained using 80% dataset and then it was further trained with 90% of data set. For these two networks' respective R values for data validation were 80% and 82% respectively. The World Health Organization's AirQ + programme was used to evaluate the effects of PM2.5 levels on health. The mean PM2.5 over the 5-year study period was 121.462(p,g/m3), about twenty-four times higher than the WHO guideline. However, if we compare the annual mean of PM2.5 concentration during the 5-year study it was shows that from 2015 the concentration dropped nearly by 8%. Out of which maximum and minimum annual mean concentration of PM2.5 was observed in 2016 and 2019 respectively. This fluctuating pollutant concentration led to maximum number of deaths 51228 in 2016 and minimum no of deaths 44920 in 2019 for all natural cases (adults age 30+ years). Additionally, a positive association between PM2.5 concentration, temperature, and wind speed were discovered. Considering the significance of predicting PM2.5 concentration for accurate and timely decisions plus the accuracy of ANN used in this study, the ANN can be utilized as an effective instrument to reduce health and economic repercussions.