Facie classification of X-field in the Niger Delta using machine learning

The Niger Delta's complex geology presents challenges for accurate facies classification and reservoir characterization. This study applies machine learning techniques to well log data from four wells in the X-Field: Freeman-001, Freeman-003ST1, Freeman-004ST1, and Freeman-005. Key features analyzed include Gamma Ray (GR), Sonic Transient Time (DT), Bulk Density (RHOB), and Neutron Porosity (NPHI). Unsupervised learning methods (K-means, GMM, and Agglomerative Clustering) yielded moderate clustering performance, with K-means achieving the highest accuracy of 47.87%, followed by GMM (20.38%) and Agglomerative Clustering (13.94%). Supervised learning methods significantly outperformed unsupervised ones. Random Forest achieved 87.06% accuracy, 87.07% precision, and an F1-score of 87.06%. Gradient Boosting delivered the best results, with 87.53% accuracy, 87.55% precision, and an F1-score of 87.54%. SVM showed lower performance, with 77.86% accuracy and an F1-score of 77.82%. These findings highlight the superior performance of supervised learning, particularly tree-based models, in facies classification and underscore the potential of machine learning to enhance reservoir characterization in the Niger Delta.