Permeability Prediction of Well Log Responses Using a Local Linear Model Tree Based on a TSK-fuzzy System

Abstract

A learned model obtained from a data set can be used for prediction, simulation, optimization, and analysis of the system. The local linear model tree used for the TSK-fuzzy system has been successful in predicting the permeability of a typical Iranian oilfield rock. The methodology involves the heuristic search to choose the space of the input partitions by axis-orthogonal splits and also automatic permeability estimation from digitized data (well logs) obtained from oil wells. The results show that the local linear model tree is incurred to the smallest error on the unseen data when compared to similar algorithms. On the other hand, the limited flexibility of the local estimation reduces the variance error due to the bias/variance dilemma. Consequently, the local estimation approach has a regularization effect on the estimation.