Design An Efficient Machine Learning Regression Models to Predict Precise CO2 Data in Terms of Global Temperature

Abstract

Global climate change is a challenge around the globe. Therefore, the increase in temperature directly correlates with the increased CO₂ emissions in environments. More CO2 emission can alter the chemical composition of atmosphere that may pose a threat to health. This paper has investigated in three iterations. The 5th order polynomial regression model is first validated on the global CO₂ anomaly dataset and on the mathematical model between CO₂ and Temperature. In the second pass, the above model is applied to the Delhi city temperature climate dataset to estimate CO₂ of the city. However, due to significant temperature difference in global dataset and Delhi local dataset. Hence, in the third and final step, it is proposed one that has temperature normalization and the process is modified. To compare the performance of Support vector Machine Regression (SVR), Gaussian Process Regression (GPR), Neural Network (NN) Regression models on CO2 Prediction. It is observed that the DVR model overall performed well with the best R2 value of 0.9397, RMSE = 7.7892 ppm and MAE = 6.4442 ppm. The CO₂ concentrations it predicts for Delhi lies between 322.269 ppm and 419.513 ppm with a mean of 379.212 ppm, a realistic range that confirms the applicability of the model beyond the global training data. It was found that the projected CO₂ contents for Delhi disclosed realistic distribution range.