International Conference on Computer and Knowledge Engineering

Home / 15th International Conference on Computer and Knowledge Engineering

Balanced Learning with Optimized Extra Trees Classifier for Reliable Lithology Identification in Imbalanced Well Log Data

Authors :

Ali Daneshpour¹ Behnam Yousefimehr² Mehdi Ghatee³

1- Department of Mathematics and Computer Science, Amirkabir University of Technology, Tehran, Iran 2- Department of Mathematics and Computer Science, Amirkabir University of Technology, Tehran, Iran 3- Department of Mathematics and Computer Science, Amirkabir University of Technology, Tehran, Iran

Keywords :

Lithology Identification،Extra Trees Classifier،Machine Learning،Well Logs،Imbalanced Data

Abstract :

Accurate lithology identification is critical for subsurface characterization in hydrocarbon exploration, yet conventional methods often fail to capture complex nonlinear relationships in well log data. To address this challenge, we propose a robust machine learning framework based on an optimized Extra Trees Classifier, enhanced by a hybrid resampling strategy to mitigate severe class imbalance. Our approach combines random oversampling of minority lithologies (e.g., coal and dolomite) with strategic undersampling of dominant classes, ensuring balanced representation while preserving critical geological patterns. Hyperparameter tuning via optimization further refines model performance, achieving an accuracy of 83.91% with a penalty score of -0.4087, demonstrating superior reliability, particularly for underrepresented facies. A comparative computational analysis confirms our framework’s efficiency, outperforming complex models such as GrowNet, Blender, and deep neural networks in both speed and scalability. To promote reproducibility, we provide the complete implementation, including preprocessing scripts and trained models, at https://github.com/alidaneshpour/ICCKE-2025