STOCK PRICE FORECASTING BASED ON COMBINATION METHOD OF BACKPROPAGATION NEURAL NETWORK AND MARKOV CHAIN

Abstract

In this era, investment in the stock sector has become one of a vertible gold mine for a community especially for investor, to invest their money. Investing in the stock field seems promising to gain a lot of profits, but any investment always have a followed risk, that is a loss obtained when making a wrong decision. Forecasting stock price using combination method of backpropagation neural network and Markov chain provides a new reference in the stock investment field. The neural network is based on gradient descent method by minimizing the total squared errors between output values and desired values. The method is based on additional momentum and adaptive learning rate. Then, discrete time Markov chain is applied in the forecast results of this method. By using combination of both methods, 80% of 20 results of the BBCA stock prices show a decreasing error rate compared to the single method of backpropagation neural network. Keywords: Backpropagation Neural Network, Markov Chain, Stock.