Perbandingan Algoritma Backpropagation Levenberg-Marquardt , Conjugate Gradient Polak Ribiere, Dan Bayesian Regularization Dalam Memprediksi Penyakit Anemia

Abstract

Lack of blood or anaemia is a condition in which the red blood cells are not functioning correctly. Therefore, the body does not get enough acid, and people with anaemia become pale and tired quickly. This study aimed to compare estimates of anaemia using the Levenberg Marquart algorithm, Polak Ribiere Conjugate Gradient, and Bayesian Regularization. Predictive information on anaemia is obtained from the Kaggle website, which consists of 1421 records. The characteristics used to calculate prediction comparisons in anaemia comprised six factors, namely Gender, Hemoglobin, MCH, MCHC, MCV and Result. Comparing the three backpropagation methods created the best architecture, namely 5-10-1, with a testing MSE of 0.0963 using the Levenberg Marquardt method.