Genetic algorithm applied to paroxysmal atrial fibrillation prediction

Abstract

Paroxysmal Atrial Fibrillation (PAF) prediction viability is an open research line. The definition of new valid parameters for this task can be based on very heterogeneous features. Genetic Algorithms (GAs) automatically find a set of parameters to maximize the diagnosis capabilities of a scheme based on the K-nearest neighbours algorithm. This is an efficient way of generating a number of possible solutions for the problem of PAF prediction. The present paper illustrates how GAs1 rather than a statistical study of the database can be used to select the parameters giving the best classification rates. © Springer-Verlag Berlin Heidelberg 2003.