Regresar

A study on the parallelization of moeas to predict the patient’s response to the onabotulinumtoxina treatment

Abstract:

This work deals with the decrease of the computational cost in the task of feature weighting for the predictive models of the response to the treatment of migraine with OnabotulinumtoxinA (BoNT-A). More specifically, we consider the multiobjective evolutionary algorithms (MOEAs) that support parallelization. All this with the aim of improving the training times of predictive models of response to the treatment. The results obtained show that accuracies close to 84 % are obtained while training times are decreased from 8 to less than 2 hours when using 8 threads. All in all, this work remarkably reduces the feature weighting execution time in comparison with Simulated Annealing, while getting similar values of accuracy.