Feature Selection in Classification of Blood Sugar Disease Using Particle Swarm Optimization (PSO) on C4.5 Algorithm

Abstract

Diabetes Mellitus (DM) is a disease caused by blood sugar level increased were higher than the maximum limit. Food consumed tends to contain uncontrolled sugar which could cause the drastic increase of blood sugar level. It is necessary to efforts, to increasing the public awareness to controlling blood sugar and the risks of increasing blood sugar level so as to determine of preventive and early detection measures One of used of data mining technique is information technology in the health sector which used a lot as a decision maker to predicting and diagnosing a several disease. This research aims to optimizing the features on classification of the data mining with the C4.5 algorithm using Particle Swarm Optimization (PSO) to detect the blood sugar level in patient. The dataset used is the effect of physical activity to the Blood Sugar Level at H. Abdul Manan Simatupang Kisaran Regional Public Hospital. The amount of dataset used is 42 record with 10 attributes. The result of this research obtained that the Particle Swarm Optimization (PSO) may increasing the accuracy performance of C4.5 from 86% to 95%. Whereas the evaluation result of the AUC Value increasing from 0,917 to 0,950. From those 10 attributes which are then selection with using PSO into 7 attributes used to determine the prediction of sugar level. Therefore the Algorithm C4.5 using the Particle Swarm Optimization (PSO) may provide the best solution to the accuracy of detection blood sugar levels.