Regresar

Comparative analysis of supervised learning algorithms for the detection of falls

Abstract:

The nature and conditions of the elderly make it prone to diseases and situations where their physical integrity may be affected; where falls are one of the highest risk factors. In this article as a main contribution, an analysis is carried out on the effect of the reduction of the space of characteristics in the classification process through the application of the Pearson Matrix. For this purpose, a comparative analysis based on metrics of 3 algorithms is presented: naive bayes, vector support machines and neural networks in the detection of falls. The signals used are accelerations in three axes obtained from the database of the Institute of Communications and Navigations corresponding to samples of 16 male and female subjects between 23 and 50 years old. The results show that naive bayes has the best performance considering the reduction in the characteristics.