The early detection system of pulmonary tuberculosis disease using learning vector quantization 2 (lvq2)

Abstract

Tuberculosis is an infectious disease caused by the Mycobacterium tuberculosis virus and infects the pulmonary. While in Indonesia, Central Java province was on the third-ranked for the highest number of new case of pulmonary tuberculosis disease. This disease can cause dangerous complications until death if not immediately detected and not treated completely. To help the community do early detection of pulmonary tuberculosis disease easily, this research aims to make an early detection system of pulmonary tuberculosis disease using Artificial Neural Network algorithm Learning Vector Quantization 2 (LVQ2). The variable that was used consisted of 8 symptoms of pulmonary tuberculosis disease. The research data obtained from health record data of pulmonary tuberculosis patients at Puskesmas Karangawen II Kab. Demak as much as 80 data. The distribution of training data and testing data was obtained from the application of k-fold cross-validation with the value of k = 8. The results showed that the best LVQ2 architecture for early detection system was obtained in combination of parameters learning rate (α) 0,06; smallest learning rate 0,001; window (ϵ) 0,3; and maximum epoch 500. The best architecture in this research produced 87,5% accuracy, 12,5% error rate, 85% sensitivity, and 90% specificity with a processing time of 8-fold was 60,68 seconds.