Getting Started With E-Journal

Lestarikan Ilmu Dengan Menulis

PENERAPAN ALGORITMA MULTILAYER PERCEPTRON UNTUK DETEKSI DINI PENYAKIT DIABETES


Ahmad Setiadi
AMIK BSI Karawang

ahmad.ams@bsi.ac.id

Paradigma Vol. XIV. No. 1, Maret 2012, ISSN:1410 – 5963

Abstract:

Each year, patient of diabetes mellitus is increasing, so that is needed the diagnose techniquewhich is effective to detect in early. Neural network as a model of data mining can be used topredict wether someone is suffered from diabetes mellitus or not. In this research, MultilayerPerceptron (MLP) as a neural network algorithm is used, not only because this algorithm has agood ability in predicting but also because this algorithm is commonly used. In this research, theprocessed data is total of 768 records and as a result of checking up Indian Pima women at least21 years old.. To implement the MLP algorithm, SPSS Neural Network 17.0 is used. The result ofimplementing algorithm then is evaluated by using confusion matrix method and ROC (ReceiverOperating Characteristic) curve method. This result is proved that implementation of MLPalgorithm to detect diabetes mellitus for Indian Pima has a good performance. The value ofaccuracy by confusion matrix method is 77,7 %. Using ROC curve method, this research showsthe accuracy of 0,83, so that it is including as good classification because it is being among 0,8until 0,9. This research proved that MLP Algorithm can be used to detect diabetes mellitus in earlytime.

Keyword:

diabetes mellitus, neural network model, multilayer perceptron, confusion matrix, kurva ROC


PDF