Forecasting Social Cost of Carbon with Predictive Machine Learning Models for CO₂ and GHG Emissions

Purpose: The purpose of this study is to ascertain the connection between oil and gas businesses' carbon emissions and the upkeep of their carbon emission monitoring systems. To forecast the carbon emissions in India, China, and all of Asia, we also want to build a machine-learning model utilizing data on carbon emissions from 1850 to 2021. This study analyses data from “Our World In Data” over a 170-year span to develop a model that forecasts future carbon emissions in Asia, particularly in India and China.

Design/methodology/approach: After extensive research, we collected data on carbon emissions, monitoring, and statistics from various sources, covering the period from 1850 to 2021. Utilizing SKLearn and Scikit-Learn libraries, we pre-processed the dataset, replaced NaN values, and split it into test and training sets (80% and 20%). Employing linear regression and decision tree regressor algorithms, we forecasted future CO2 and GHG emissions in China, India, and Asia. The decision tree model outperformed linear regression in predicting emissions accurately. Additional data visualization aided in understanding actual versus predicted emissions. Findings: 95% is the selected level of confidence. The logistic regression model indicates that p is bigger than alpha. Considering this, we adopt the null hypothesis. The difference between actual and expected data is not very significant, according to the chi-square (2) value of 0.3241, which assesses the disagreement between observed and predicted frequencies of outcomes of a set of events or variables. We concluded that the outcomes produced by the decision tree regressor algorithm were superior and less error-prone than those obtained by linear regression.

Originality/value: Machine learning model for forecasting CO2 and GHG emissions. Hypothesis formulation and using logistic regression to accept the null hypothesis: Companies monitoring carbon emissions emit less carbon.