CLASSIFICAÇÃO COM ÁRVORES DE DECISÃO EM PARALELO

Abstract

This work presents a synthesis of the functioning of a data classification algorithm, which classifies items or samples according to the characteristics acquired through training carried out on a database. These trainings are done using decision tree models as a tool to analyze the data. The classification process can be applied when the objective is to predict future occurrences, such as for a company, in improving the sales plan or in the relationship with customers. The project resulted in the demonstration of the functioning of a serial and a parallel classifier, presenting its methodology, structure, the mathematical model that composes it and its applications. A comparison was also made of these two different perspectives of software construction (serial and parallel), noting that, in the parallel algorithm, there was a considerable reduction in execution time compared to the serial and showing when this gain can be advantageous in an application.