Perpustakaan Digital ITB

The process of well to well correlation can be time-consuming as well log data grows massively. The conventional approach inclines to be subjective as it is based on one’s perspective on the data. In this study, an automated well to well correlation using the machine learning method was used. Furthermore, this method can be considered to be convenient, and time-saving compared to the traditional approach. A supervised learning method was used in this study, five types of logging data were used and labeled at a certain depth. Over ten thousand data points were used as an input and output of the machine learning model. 70% of the data was used to train the model while the other 30% was used to validate the trained model. Different approaches were used to create the model. The model was then tested on log data from different well to see the correlation within. Furthermore, the hyperparameter optimization was used as model evaluation to seek the best parameter. The model with the best performance was then justified as the selected model. Three models were created using the K-Nearest Neighbors, Stochastic Gradient Descent, and Multilayer Perceptron approach respectively. Overall, the K-Nearest Neighbor approach was justified as the selected model. With a cross validation score of 0.981, the model resulted in 95.35% and 87.92% of accuracy on a first test set and second test set respectively. The created model based on the machine learning approach can harness massive log data and can be used on well to well correlation in a less time-consuming manner.