Application and Comparison of Machine Learning Methods for Mud Shale Petrographic Identification

Machine learning is the main technical means for lithofacies logging identification. As target of shale oil spatial distribution prediction, mud petrography subjected to constraints stratigraphic inhomogeneity and information redundancy. Therefore, choosing most applicable machine method different geological characteristics data situations one key aspects high-precision However, only a few studies have been conducted on applicability methods petrography. This paper aims identify using commonly used methods. The study employs five supervised algorithms, namely Random Forest Algorithm (RF), BP Neural Network (BPANN), Gradient Boosting Decision Tree Method (GBDT), Nearest Neighbor (KNN), Vector (SVM), as well four unsupervised K-means, DBSCAN, SOM, MRGC. results are evaluated confusion matrix, which provides accuracy each algorithm. GBDT algorithm has better in learning, while K-means DBSCAN algorithms higher learning. Based comparison it can be concluded that identification poses challenges due limited sample high overlapping degree type areas. selecting appropriate crucial. Although generally accurate, they by volume samples. Future research should focus how make samples combine explore types non-coring wells.