The transition to renewable energy sources (RES) is crucial for sustainable development and climate change mitigation, particularly in developing countries with rapidly increasing energy demands. However, most existing prediction methods for solar energy adoption, often based on linear regression and basic statistical techniques, fail to capture the intricate interactions among socio-economic, environmental, and policy factors. Consequently, these methods can lead to suboptimal predictive accuracy and provide limited actionable insights for policymakers. This study employs advanced machine learning techniques to predict solar energy adoption in developing regions, utilizing historical data from 2000 to 2021. By integrating key socio-economic, environmental, and policy features, we develop a predictive model using the ExtraTreesRegressor, optimized through feature engineering and hyperparameter tuning via RandomSearchCV. Our results highlight non-renewable electricity consumption, GDP, and solar electricity consumption per capita as the most influential factors. Conversely, variables such as population, year, per capita electricity consumption, and annual CO2 emissions (per capita) have minimal impact. The model demonstrates high predictive accuracy, achieving an R2 value of 0.987 and an MSE of 0.390 on the test set, and an R2 of 0.907 and an MSE of 4.056 on cross-validation. These results highlight the importance of advanced machine learning techniques and comprehensive data integration in improving model performance. By providing a detailed analysis of key determinants, this research offers valuable insights for policymakers and stakeholders, facilitating a smoother transition toward sustainable energy futures in these critical regions.