A Bio-inspired Fusion Method for Data Visualization

Abstract

This research presents a novel bio-inspired fusion algorithm based on the application of a topology preserving map called Visualization Induced SOM (ViSOM) under the umbrella of an ensemble summarization algorithm, the Weighted Voting Superposition (WeVoS). The presented model aims to obtain more accurate and robust maps, also increasing the models stability by means of the use of an ensemble training schema and a posterior fusion algorithm, been those very suitable for visualization and also classification purposes. This model may be applied alone or under the frame of hybrid intelligent systems, when used for instance in the recovery phase of a case based reasoning system. For the sake of completeness, the comparison of the performance with other topology preserving maps and previous fusion algorithms with several public data set obtained from the UCI repository are also included.