Lithology Recognition by Neural Network Ensembles

This paper investigates the advantages of methods based on Neural Network Classifier Ensembles - sets of neural networks working in a cooperative way to achieve a consensus decision- in the solution of the lithology recognition problem, a common task found in the petroleum exploration field. Classifier ensembles (Committees) are developed here in two stages: first, by applying procedures for creating complementary networks, i.e., networks that are individually accurate but cause distinct misclassifications; second, by applying a combining method to those networks outputs. Among the procedures for creating committee members, the Driven Pattern Replication (DPR) was chosen for the experiments, along with the ARC-X4 technique. With respect to the available combining methods, Averaging and Fuzzy Integrals were selected. All these choices were based on previous work in the field. This paper proves the effectiveness of applying ensembles in the recognition of geological facies and suggests algorithms that might be successfully applied to others classification problems.