Predicting the Oil Investment Decision through Data Mining

Abstract

Petroleum investment decision remains subject to economic and financial research for decades. Due to capital intensive and higher risk on oil exploration, the investment decision has become more important than ever before.  This study aims to shed some light on this issue by conducting four machine learning algorithms to predict the decision applying the dataset from 2007-2019. This study includes the Decision Tree, Random Forest, Naïve Bayes, and Support Vector Machine. A comparative performance analysis is the illustrated using confusion matrix, Cohen’s Kappa value, and the accuracy of each model and Area under the ROC Curve. In this study, a machine learning approach was carried out on the oil exploration data. The findings demonstrate that Naïve Bayes has the most accurate performance for the classification (94.5%), followed by Decision Tree (92.9%), Random Forest (90.9%), and Support Vector Machine (89.6%). In practice, the selected Naïve Bayes model was applied to assess the decision using a new data test. The findings can diminish the subjective blindness and confirmation bias in the investment decision and bring about a reasonable and orderly exploration and development of petroleum reserves.