Accurate estimation of petrophysical properties from well log data is paramount for reliable reservoir characterization and informed decision-making in hydrocarbon exploration and production. Conventional methods often struggle with the inherent complexities and non-linear relationships within geological datasets, leading to suboptimal prediction accuracy. To overcome these limitations, we propose a novel hybrid machine learning method that integrates multiple predictive models with a unique residual adjustment strategy. A novel aspect of the methodology involves fitting a Weibull distribution to the residuals of machine learning models, such as random forest, support vector regression, and artificial neural networks as foundational learners. A distinctive aspect of this approach is the application of Weibull distribution analysis to model and subsequently adjust the residuals generated by these individual base models, thereby enhancing individual model predictive accuracy. These adjusted base models were then combined into a stacked ensemble, utilizing a ridge regressor as the final meta-learner to further consolidate their predictive strengths. Performance evaluation, conducted using metrics such as R-squared (R²) and root mean squared error (RMSE), demonstrated that the proposed stacked ensemble model significantly outperformed individual models, achieving a superior predictive capability for the photoelectric factor. The integration of residual analysis through Weibull distribution further contributed to the overall predictive robustness. This research demonstrates the efficacy of advanced ensemble machine learning techniques, particularly when combined with detailed residual distribution analysis, in accurately characterizing complex subsurface properties. The developed method offers a powerful and reliable tool for enhancing reservoir modeling and supporting more effective decision-making in geological applications.