PEMODELAN GENERAL REGRESSION NEURAL NETWORK (GRNN) PADA DATA RETURN INDEKS HARGA SAHAM EURO 50

Abstract

General Regression Neural Network (GRNN) is one of the network models that is used for the radial basis function approach. GRNN models including neural network models with a quick solution, because it does not need a large iteration in estimation weights. This model has a standard network architecture, where the number of units in the pattern layer in accordance with the amount of input data. One application GRNN is to predict stock returns of the index value of Euro 50 CFD (Contract for Difference). Euro 50 Index CFD (Contract for Difference) is used as the benchmark stock price of the 50 largest companies in the euro zone. The investors to invest in the stock index Euro 50 CFD (Contract for Difference) with expectation of obtaining appropriate rewards back to what has been invested in. GRNN models showed that the value of RMSE and R2 for training data and 0.00095 and 99.19%. For testing the data obtained RMSE and R2 value of 0.00725 and 98.46%. Based on the forecast value of the stock return next twelve days obtained the highest loss or capital loss on December 10, 2014 at 5.583188% and the highest profits or capital gains on December 10, 2014 by 2.267641% Keywords: GRNN, Neural Network , Stock Return, Euro 50, Forecasting, Capital Loss, Capital Gain, Forecasting