Pemanfaatan Algoritma Fletcher-Reeves untuk Penentuan Model Prediksi Harga Nilai Ekspor Menurut Golongan SITC

Abstract

Conjugate gradient Fletcher-Reeves algorithm, according to some literature, is an optimization method that is suitable when juxtaposed with the backpropagation method because this method can speed up the training time to achieve a minimum convergence value. Therefore, this study aims to prove whether the algorithm has good performance and can provide efficient convergence results when used to solve prediction problems in the case of export values ​​according to the Standard International Trade Classification (SITC) class. The results of this study are a predictive model that can be used and developed to make predictions in seeing the development of the export value of the SITC class based on the US Dollar currency. The research data was taken from the website of the Central Statistics Agency for 2010-2020. Prediction models that will be analyzed using the Fletcher-Reeves algorithm include 5-20-1, 5-25-1, and 5-30-1, with the activation functions of tansig and logsig. Based on the analysis carried out through excel calculations from the training and testing process using the Matlab-2011b application, the results obtained that the 5-25-1 network model is the best model with a performance value or Mean Square Error 0.00287273 compared to the other four models. So it can be concluded that the Fletcher-Reeves algorithm is proven to produce faster convergence; it can be seen from the epoch generated from each model that it is not too large and the time required is relatively short