Publications

Facies classification using well log data is an important process in oil and gas exploration, which enables us to understand the reservoir character and the concept of the petroleum system in the target area. Time consuming practices and biased results from conventional method of identifying facies by using human interpretation are still greatly used today. With the evolving era of Industrial Revolution 4.0, there have been rapid growth of cognitive automation and computation in the field of geoscience. Machine learning applications are often used in geoscience data processing methods, as identification of lithology facies from well log data being one of them. Using the machine learning concept, lithofacies classification activities using well logs can be done more effectively. It will lead to increased productivity in oil and gas exploration activities. Nonetheless, there are various machine learning algorithm that can be used, which comes with its own characteristics. This paper aims to analyze the characteristics of several machine learning algorithm in recognizing and evaluating the lithology patterns of the well logs. Data model used in this research are open source well log data from Hugoton and Panama oil field that refers to case studies of sandstone and carbonate HC reservoirs in Indonesia. These models are then compared and observed in five machine learning algorithms; Na�ve Bayes, Random Forest, Support Vector Machine (SVM), K Nearest Neighbors, Gradient Boosting and Decision Tree. Comparison of these algorithms will present its advantages and disadvantages, depending on the reservoir case. It will help to find the suitable machine learning algorithm for each unique reservoir case. The Gradient Boosting algorithm is the best one to help geologist determine sandstone reservoir.