(1901-4946) On the use of Heronian means in a similarity classifier

Abstract

This paper introduces new similarity classi fiers using the Heronian mean, and the generalized Heronian mean operators.We examine the use of these operators at the aggregation step within the similarity classifi er. The similarity classifi er was earlier studied with other operators, in particular with an arithmetic mean, generalized mean, OWA operators, andmany more. The two classi fiers here are tested on four real world data sets, i.e., echocardiogram, fertility, horse-colic, and lung cancer. Three previously studied similarity classi fiers are used as benchmarks to the new approaches. Weobserve that the similarity classi er with a generalized Heronian mean produces good classi cation results for the tested data sets, and is therefore more suitable for use in these classi fication problems.