Model neural network proses nonlinear autoregressive data finansial Studi kasus : data indeks harga saham gabungan bursa efek Surabaya = Neural Network Models for Nonlinear Autoregressive Process at Financial Data ...

This paper discussed about the neural network models at nonlinear autoregressive process which is applied in the Composite Stock Price Index data at Surabaya Stock Exchange. One of the problem in fitting NN models is that an NN models which fits well may give poor out-of-sample forecasts. Thus we think it is unwise to use traditional modeling skills to select a good NN model, e.g. to select appropiate lagged variables as the `inputs'. The tests of linearity needed to decide that the data agree to be predicted with nonlinear models like NN. The Pruning methods which belong to the general-to-specific procedure is used to choose the optimal number of hidden units. The size and topology of the used networks is found by reducing the size of the network through the use of multiple correlation Coefficients and graphical analysis of network output per hidden layer cell. The results of NN models for various architecture are compared with those obtain from the Box-Jenkins methods. The Akaike's Information Criterion, Schwartz Bayesian Criterion and Mean Squate Error are used for comparing different models. Keywords : nonlinear autoregressive, neural network, linearity, pruning, composite stock price inde