Perpustakaan Digital ITB

This study provides workflow to predict porosity and permeability by adding facies and formation label to machine learning algorithm. Limited core sample analysis and manually interpret facies label problems are the motivation to provide this workflow as alternative solution. Start from creating robust supervised machine learning model for predicting formation and facies label using well log data with feature augmentation to improve model. Each predicted values will be predictor for next target variable in which making integrated workflow. Implementing proposed feature augmentation by (Bestagini, Lipari, & Tubar, 2017) and (Marcelo, 2018) to “X” field log data and core analysis data gives 0.91 f1 score in classifying formation and 0.78 f1 score in classifying facies. Prediction of porosity using added features from facies and formation label gives 0.804 R2 score. Permeability prediction by transforming train target variable to log values not sufficiently enough to accurately predict permeability since it gives only 0.175 R2 score.