Missing values absorption based on homogenous granulation

Abstract

In the recent works we have developed new methods of granulation-homogenous variants-in the family of techniques founded by Lech Polkowski. The level of approximation of decision systems for these methods is not the highest among previously developed algorithms, but the basic advantage is its dynamic behaviour-there is no need to estimate any parameters for this methods. Granulation level is dependent on the indiscernibility ratio of decision classes. Additionally to the fact that our new methods found successful application in the context of data approximation is their utility in ensemble models. We have developed a novel technique-ensemble of random granular reflections. In this paper we continue the series of works and we have carried out experiments to check effectiveness of our new granulation technique in the context of missing values absorption.