MACHINE LEARNING APPLICATION FOR CARBON ESTIMATION – A CASE STUDY

Abstract

Abstract. Climate change is a most global challenging issue. In this regard, a study on carbon dioxide, one of the pollutants causing the climate change is demonstrated in two different states of India viz., Visakhapatnam district (AP) and Shastri Nagar (RJ) for the period of April 2022 to January 2023. Carbon dioxide (CO2) is experiential on hourly and monthly basis for different seasons – summer (April), rainy (July) and winter (December). Most of air pollutants include NO2, CO2, PM2.5, PM10 etc. that are the major cause for climate change. The air quality in these zones is very poor, highly polluted and risk to humans. The study proved that CO2 is found comparatively low in rainy season over other. The machine learning regression models were modelled for Visakhapatnam and best models obtained are 1. Step-wise Linear Regression model with MSE (4.51E-28), RMSE (2.12E-14) and R-Squared (1) identified for rainy month. 2. Neural Network Narrow model with MSE (0.462), RMSE (0.680) and R-Squared (0.999) for winter month. 3. Linear Regression model with MSE (0.108), RMSE (0.329) and R-Squared (0.999) for summer. Similarly, the best models for Shastri Nagar for monthly data are 1. Step-wise Linear Regression with RMSE (20.292), MSE (411.774) and MAE (12.524) for April & May (summer), 2. Neural Network Narrow model had RMSE (3.399), MSE (11.554) and MAE (2.141) for July (rainy). 3. Neural Network Bilayered model with RMSE (1.618), MSE (2.619) and MAE (0.593) for November & December (winter). The results obtained were very efficient and reliable.