Discretization of time series dataset with a genetic search

Abstract

In this work we propose a new approach to the discretization of time series using an approach that applies genetic algorithm operations called GENEBLA. The basic idea is to minimize the entropy of the temporal patterns over their class labels, follow a genetic search approach that allows to find good solutions more quickly to explore a wide variety of possible ways to solve the problem at the same time. The performance of GENEBLA was evaluated using twenty temporal datasets and compared to an efficient time series discretization algorithm called SAX and EBLA3 algorithm that shows similar representation. © 2009 Springer-Verlag Berlin Heidelberg.