ANÁLISE DE SÉRIES NUMÉRICAS UTILIZANDO APRENDIZADO NÃO SUPERVISIONADO

Abstract

Numerical series analysis is a task that brings a great demand of time and cost and, therefore, it becomes necessary tools that speed up and reduce the cost of these analyzes. The respective work presents the design of a classification model for the set of numerical series, thus enabling the automation of numerical series analysis, especially in series that have a large number of elements, as one of the phases of the model culminates in the compression of series. It was found that the classifier has low performance in numerical series, as they have few elements, however it was found that the performance was high in numerical series, with many elements. From these results we can conclude that, although the classifier's performance is low in small series, in the context of several areas they generate data in the form of numerical series, the classification model would bring great benefits, as a decision support tool, in several areas.