Impact of data particle divide depth level on effectiveness of hypergeometrical divide classifier

Abstract

Gravitational classifiers belong to the supervised machine learning area, and the basic element they process is a data particle. So far, many algorithms have been presented in the world literature. They focus on creating a data particle and determining its two important parameters – a centroid and a mass. Hypergeometrical divide is one of the latest algorithms in this group, which focuses on reducing the amount of processing data and keeping relevant information. The proportion of data to information depends on the data particle divide depth level. Its properties and application potential have been researched, and this article is the next step of the work. The research described in this article aimed to determine the relation of the depth level value of data particle divide to the effectiveness of the hypergeometrical divide algorithm. The research was conducted on 7 real data sets with different characteristics, applying methods and measures of evaluating artificial intelligence algorithms described in the literature. 63 measurements were performed. As a result, the effectiveness of the hypergeometrical divide method was defined at each of the available data particle divide depth levels for each of the used databases.