Gandhi Rifal, Hizir Sofyan, Romi Satria Wahono
Accurate lithology prediction is critical for subsurface modeling in oil and gas exploration. However, while machine learning (ML) techniques have been applied to automate lithology classification, class imbalance and noisy attributes are a challenge. This study proposes a novel integration of SMOTE, Evolutionary Algorithm, AdaBoost and K-Nearest Neighbor (KNN) to boost the performance of lithology classification. To this end, SMOTE is employed to tackle the problem of imbalanced dataset while Evolutionary Algorithm (EA) is implemented to feature selection to alleviate the influence of noisy attributes. Also, AdaBoost increases the model robustness by repeating learning process focusing on misclassified samples. The proposed method is tested on the publicly available FORCE 2020 Well Log and Lithofacies Dataset, which contains measurements from multiple wells and comprises eleven lithology classes. It is found that the hybrid approach achieves better performance than the standalone KNN method, attaining 96.75% accuracy and an 89.67% F1-Score. The results of the study reveal that incorporating data balancing, feature selection, and boosting techniques enhances the lithology classification, solving the problems typical of well-log data. © 2025 IEEE.
Universitas Syiah Kuala, Informatics Department, Banda Aceh, 23111, Indonesia; Universitas Syiah Kuala, Statistics Department, Banda Aceh, 23111, Indonesia; Brainmatics Indonesia Cendikia, Jakarta, 12870, Indonesia
Research at a Glance
Register to unlock
Topics & SDG Alignment
Register to unlock
Collaboration
Register to unlock
Author Profile (Selected)
Register to unlock
References Overview
Register to unlock
Journal & Source
Register to unlock
Metadata & Integrity
Register to unlock