Revisão de métodos para análise de agrupamento de dados em data mining

Abstract

The core components of data mining technology have been in development for decades. Today, the maturity of these techniques, coupled with high-performance database engines and extensive data integration efforts, make these technologies practical for today's environments. The cluster analysis aims to separate objects into groups, grouping them according to their characteristics in common with a predetermined criterion, identifying comprehensible patterns. To perform this classification the various data mining techniques use complex mathematical functions. In this context, even an easier abstraction of the formulas for grouping data is not simple to understand, especially for those who are not from the area or have no knowledge of mathematical concepts. The purpose of this work is to clarify the formulas of some methods of grouping data, explaining them in a practical and objective way, with examples, of how they work. For this, 3 algorithms of the same genre, k-means, k-medians and k-medoids were chosen to be detailed using the same set of data.