Feature selection-Based ANN for improved characterization of carbonate reservoir

Permeability of hydrocarbon reservoir is an important petrophysical parameter that serves as an indicator of the overall quality and quantity of hydrocarbons present in the reservoir and the rate at which it can be produced. Therefore, its accurate prediction is of fundamental concern to petroleum engineers. In this work, a correlation-based feature selection technique is proposed to improve the performance and accuracy of artificial neural network (ANN) in permeability prediction. The effect of the technique has been investigated using two diverse datasets obtained from a Middle Eastern oil and gas field. The proposed approach employs fewer datasets in substantially improving ANN performance. The results of this work suggest a way to improve the performance of computational intelligence technique in reservoir characterization using fewer datasets which results in less computing time and computational cost.