Performance of Algorithms for Interval Discretization of Biomedical Signals

Abstract

A methodology to quantify the dependence be tween features using the Ameva discretization algorithm and the advantages of qualitative models is presented in this paper. This approach will be applied over medical data sets. A com parison among Ameva and other related works has been done. The results, as will be depth explained in this paper, show that Ameva-based methodology can be used to determine the depen dence between features in a fast and understandable way from data sets with a high number of attributes and low number of instances. This is a quite important feature in genomic en vironments among others. This methodology has been applied to some well-known medical data sets and the results obtained shown that is a good alternative to other established algorithms in terms of clarity and computational cost.