The Combination of C4.5 with Particle Swarm Optimization in Classification of Class for Mental Retardation Students

Abstract

Mental retardation or brain weakness is a condition of children who experience mental disorders. There are several characteristics to know the child has mental retardation. When entering a school, teachers are expected to be able to determine the right class for mental retardation students according to their category. Data mining is the process of finding patterns in selected data using artificial intelligence and machine learning. Algorithm C4.5 is one of the classification techniques in data mining. C4.5 can be used to create decision trees and classify data that has numeric, continuous, and categorical attributes. But C4.5 has the disadvantage of reading large amounts of data and cannot rank every alternative. PSO is an optimization algorithm for feature selection that can improve performance in data classification. Therefore, this study proposes an algorithm that can overcome the weaknesses of C4.5 by combining PSO. This study aims to classify a class of new mental retardation students using a combination of C4.5 as a classification and PSO as a feature selection to determine the attributes that affect the level of accuracy. The contribution of this research is to make it easier for the school to determine the new class of mental retardation students so that it is appropriate and according to their needs. The classification process in this study uses a combination of C4.5 and PSO. The validation used in this model is 10-fold cross-validation, and the evaluation uses a confusion matrix. This study resulted in an accuracy of C4.5 before using PSO of 91%. While the accuracy of C4.5 uses a PSO of 93%. Of the 20 attributes, there are 6 attributes that affect the level of accuracy. This study shows that PSO can be used to implement feature selection and increase the accuracy value of C4.5 by 2%.