Abstract
In this recent years, digitalization and automation have been massively transforming global industries to a new era of technology. This also applies in the petroleum industry by utilizing machine learning to aid exploration activities and analyze past data to extract the unseen data. This research implemented the usage of machine learning in petroleum geoscience by utilized the k-nearest neighbor (knn) as the supervised learning model algorithm to predict electrofacies pattern, lithofacies, and depositional environment classification using p-2 and k-1 well data as training dataset and b-1 well data as testing dataset where all the wells are on the same area. The workflow also involves the training-validation-testing method, hyperparameter tuning, and feature engineering to get the models best hyperparameter combination and score. The confusion matrix result from validation stages shows that the knn algorithm generates a quite similar prediction compared with the actual data and produces a good test score with values of accuracies above 0.75 from three prediction cases. The model also generates a good electrofacies pattern, lithofacies, and depositional environment classification results based on the b-1 well plot and indicates that there is a plover formation continuity as the reservoir zone that might be having a marine-influenced depositional environment.