Implementasi Metode Machine Learning menggunakan Algoritma Evolving Artificial Neural Network pada Kasus Prediksi Diagnosis Diabetes

Abstract

Diabetes mellitus is a global health problem that can affect anyone, from children, adolescents, to adults. Therefore, diabetes is one of the non-communicable diseases that has become a serious threat to global health. Since 1980, the number of diabetics worldwide has nearly doubled from 4.7% to 8.5% of the total population. The International Diabetes Federation (IDF) even estimates that the number of diabetes sufferers worldwide will reach 700 million people by 2045. In response to this condition, this study predicts diabetes diagnosis using machine learning algorithms, artificial neural network. However, there is a major problem with this algorithm, namely in determining the correct architecture. This problem can be viewed as an optimization problem, where many architectural possibilities that can occur. Therefore, to search for the right architecture to increase the accuracy of the predictions, there will be stages to use the evolution algorithm. Because this algorithm is very suitable to be applied in an optimization case. This study implements Evolving Artificial Neural Network (EANN) algorithm to predict the patient's diagnosis. It is with the hope that this study can produce higher accuracy in predicting patient diagnosis in diabetes. The data set used was Pima Indian Diabetes from the UCI Machine Learning Repository. Based on the experiments that have been carried out, the best model produced has an accuracy of 83.55%. This means that the algorithm used is quite successful in predicting diabetes diagnosis.