Optimasi Conjugate Gradient Pada Algoritma Backpropagation Neural Network Untuk Prediksi Kurs Time Series
Abstract
This paper investigates the use of methods of propagation neural network to forecast the rupiah against the dollar in time
series . Data is divided into two parts , namely the training data test data . Having obtained the best weight training results
data then after that test data is used to see the results of MSE . Training algorithm uses basic backpropagation gradient
descent algorithm and then optimized using the conjugate gradient . The results show the value of the basic algorithm
backpropogation MSE gradient descent and conjugate gradient of 1.02159 mengsilkan MSE of 0.0198012 . From these
results indicate conjugate gradient algorithm produces smaller error.