Evolutionary hybrid learning approach for neural networks applied to medical instances classification

Abstract

Artificial intelligence has helped several medical areas such as the field of medical diagnosis, in which it has served as a support to the health professional to issue a diagnosis. This is achieved through various techniques of artificial intelligence such as decision trees, intelligent agents, support vector machines, data mining or neural networks. One way to use neural networks to support medical diagnosis is to use it as classifier. This is, based on a historical and representative clinical file, to construct a neural model that allows to classify new cases. Evolutionary networks according to the literature represent a more precise tool than the classical neural networks. In this paper we propose the use of a doubly evolutionary hybrid neural network that evolves both, the topologies and the synaptic values of each topology. Experimentation shows that the evolutionary hybrid neural network obtains competitive results with respect to the classic evolutionary neural networks.