Independent and Simultaneous Evolution of Fuzzy Sleep Classifiers by Genetic Algorithms

This paper describes two alternative approaches to the automatic inference of a fuzzy classification system applied to computerised sleep staging. Both approaches use genetic algorithms to evolve a fuzzy classifier per sleep stage. In the first case, each stage classifier is independently evolved while in the second case, the classifiers are evolved simultaneously. Satisfactory results where obtained for the individual stage classifiers (76% to 97%), but the global performance of the classification system decreased significantly. No significant differences between the two evolution methods were observed. Possible improvements are suggested.