Evolución de un método basado en símbolos para clasificar series temporales usando minería de datos.

Abstract

The diagnostic analysis of pathologies related to the auditory system evaluated with brainstem auditory evoked potentials (BAEPs) has been performed manually by the specialist. To make up for the lack of automatic processes for the evaluation of BAEPs, a symbol-based method, Symbolic Pattern-based Classification (SPC), was proposed to support the diagnosis of the mentioned pathologies, which is based on the analysis of BAEPs converted into symbolic time series. In the present research, the algorithm for the numerical to symbolic transformation was modified using cubic splines to smooth the time series by eliminating the residual noise in the taking of these tests and the symbolic time series classification algorithm. our classification method gave very promising results: 99.4% accuracy, 97.8% sensitivity and 100% specificity, which are very similar to the results obtained with the original method, however the execution time of the implemented algorithms are much better than that of the original ones.