Publications

This study will focus on identifying depositional facies in uncored intervals using a gradient boosting classifier, based on electric logs: gamma-ray (GR), resistivity (ILD), neutron porosity (NPHI), and density (RHOB), as well as facies description and classification derived from cored intervals. Supervised learning with gradient boosting classifiers is the primary method that combines a lot of weak learning models to create a robust predictive model. A gradient boosting classifier was applied because the output will be in the form of images. We used nine wells such as four training data, and five testing data along with gamma-ray, resistivity, NPHI, and RHOB as input. The statistical methods were used to distribute facies on each well, and we used the F1 score and average of confusion matrix to validate the values. The result shows 0.718 or 71.8% of the F1 score and 0.6617 or 66.17% of the confusion matrix. With this level of accuracy, we conclude that the gradient boosting classifier methods are reliable enough to determine facies in the area that have limited core data with satisfying efficient results without reducing the accuracy.