Abstract

The accurate prediction of petrophysical properties such as permeability and porosity play a vital role in optimizing hydrocarbon exploration and reservoir characterization. This study introduces an integrated machine learning framework employing Random Forest, Support Vector Machine, and Decision Tree algorithms to predict permeability and porosity from petrophysical well log data. A high quality dataset was curated and split into training (70%), validation (15%), and testing (15%) subsets to ensure model generalization and minimize overfitting. Hyperparameter optimization was conducted using Grid Search, Random Search, and Bayesian Optimization techniques. Model performance was evaluated through key metrics including accuracy, mean squared error (MSE), R² score, and mean absolute error (MAE). The results demonstrate that the optimized Random Forest model outperformed the other algorithms in terms of accuracy and robustness. Feature importance analysis further emphasized the contribution of key geological parameters to model predictions. This research highlights the effectiveness of hyperparameter tuning in enhancing model performance and provides a robust data-driven framework for petrophysical analysis. The findings contribute to the advancement of AI-based methodologies in hydrocarbon reservoir assessment and support the use of machine learning models as efficient tools for reducing uncertainty in subsurface characterization.

Authors

Asror Boytemirov Maxmadustovich
Tashkent University of Information Technologies named after Muhammad al-Khwarizmi, Uzbekistan

Keywords

Machine Learning, Support Vector Machine, Random Forest, Decision Tree, Geological Multimodal Data, Hyperparameter Tuning