Document Type : Research Paper
1 Petroleum, Mining and Materials Engineering Department, Islamic Azad University, Central Tehran Branch, Tehran, Iran,
2 Tehran University
3 Petroleum, Mining, and Materials Engineering Department, Islamic Azad University, Central Tehran Branch, Tehran, Iran,
In this study, in order to facies classification, the kernel principal component analysis (KPCA) feature extraction method is used to extract new features from the measured well-logs. After applying the Principal Component Analysis (PCA), and KPCA feature extraction approaches, the classification was made using three powerful classifiers: Multilayer Perceptron Neural Network (MLP), Support Vector Machine (SVM), and Random Forest (RF). Finally, the predicted results for the test data that were not included in the training process were evaluated with the F1 score criterion.
The PCA method did not show a significant effect on the classification performance due to the nonlinear structure of the facies. Our results show that the KPCA improves the performance of facies classification. Compared with the conventional approach based on well-log data, our new approach improves the classification accuracy for each classifier algorithm. In the RF results, the classification accuracy has increased by about 6% while using the KPCA feature extraction approach, increasing from 52% to 58% compared to the original well-log data.