Comparison Of Facies Estimation Using Support Vector Machine (SVM) And K-Nearest Neighbor (KNN) Algorithm Based on Well Log Data

Abstract

Facies classification is the process of identifying rock lithology based on indirect measurements such as well log measurements. Usually, the facies are classified manually by experienced geologists, so it takes a long time and is less efficient. In this paper, two machine learning (Support vector machine and K-Nearest Neighbor) were adopted to increase the effectiveness and shorten the time process of facies classification in Z Field, Indonesia. The machine learning algorithm was carried out in 4 steps, i.e. data selection, training phase, verification, and validation stage. The machine learning input data are density log, gamma ray log, resistivity log, SP log; and the output facies target are Sandstone, Siltstone, Claystone, and Limestone. The data is divided into train data for the training process and test data to validate the machine learning output. In Support vector machine results, the training accuracy is 70.1% and the testing accuracy is 47.4%, while in KNearest Neighbor results, the training accuracy is 70.1% and the testing accuracy is 63.3%. This result showed K-Nearest Neighbor has better accuracy than the support vector machine in facies classification in the Z field.

Keywords

• geophysics