Binary Classification for Hydraulic Fracturing Operations in Oil & GasWells via Tree Based Logistic RBF Networks

Abstract

In this paper we develop a novel tree based radial basis function neural networks (RBF-

NNs) model incorporating logistic regression. We aim to improve the classication performance of logistic regression method by pre-processing the input data in RBF-NN frame. Although the scope of our proposed method is binary classication in this paper, it is easy to generalize it for multi-class classication problems. Furthermore, our model is very convenient to adapt for n < p  classication problem that is very popular yet dicult topic in statistics. We show the generalization and classication performance of our model using simulated data. We have also applied our model on a real life data set gathered from hydraulic fracturing in oil & gas wells. The results show the high classication performance of our model that is superior to logistic regression. We have coded our model on R software. Logistic Regression applications were carried out using IBM SPSS Version 20.