Characteristics and Potential Developments of Multiple-MLP Ensemble Re-RX Algorithm

Abstract

AbstractIn this paper, we carefully review all our work since 2012 and establish the concept of the Multiple-MLP Ensemble Re-RX algorithm. We first examine the background and procedures of the Recursive-Rule Extraction (Re-RX) algorithm family and its variants, including the Multiple-MLP Ensemble Re-RX algorithm (“Multiple-MLP Ensemble”), which uses the Re-RX algorithm as its core to find the rules for seven kinds of mixed (i.e., discrete and continuous attributes) datasets. We compare the accuracy against only the Re-RX algorithm family. The Multiple-MLP Ensemble Re-RX algorithm cascades standard backpropagation (BP) to train a multiple neural-network ensemble, where each neural network is a Multi-Layer Perceptron (MLP). Strictly speaking, multiple neural networks do not need to be trained simultaneously. Therefore, the Multiple-MLP Ensemble avoids the previous complicated neural network ensemble structures and the difficulties of rule extraction algorithms. The extremely high accuracy of the Multiple-MLP Ensemble algorithm generally outperformed the Re-RX algorithm and its variant. The results confirm that the Multiple-MLP Ensemble approach facilitates the migration from existing data systems toward new analytic systems and Big data.KeywordsRe-RX algorithmRule ExtractionMLPData MiningBig DataNeural Network EnsembleMixed DatasetEnsemble Concept