Predicting the environmental impact of financial development with machine learning algorithms

Abstract

This study aimed to determine the importance levels of variables affecting carbon emissions. Various machine learning algorithms were used and their performances were compared. Energy consumption, gross domestic product (GDP), interest rate, credit volume, inflation, and uncertainty index were used as independent variables. Model performances were evaluated using MAE, MSE, RMSE, and R2 metrics. Among the SVR, KNN, RF, ANN, XGBoost, and LightGBM algorithms, the Random Forest had the highest predictive power. The analysis revealed that the most influential variable on CO2 emissions is energy consumption, followed by GDP, interest rate, credit volume, inflation, and uncertainty index. The results emphasize optimizing energy consumption, increasing efficiency, and switching to renewable energy to reduce carbon emissions. To ensure environmental sustainability, the study recommends increasing technology incentives, prioritizing the use of renewable energy, and policymakers to develop interest and credit policies to reduce CO2 emissions. Thus, it states that economic growth can achieve a sustainable structure with environmentally friendly steps.