PERAMALAN HARGA SAHAM MENGGUNAKAN RECURRENT NEURAL NETWORK DENGAN ALGORITMA BACKPROPAGATION THROUGH TIME (BPTT)

Susanti, Linda Aqnes Desi and FARIZA, ARNA and SETIAWARDHANA, SETIAWARDHANA (2011) PERAMALAN HARGA SAHAM MENGGUNAKAN RECURRENT NEURAL NETWORK DENGAN ALGORITMA BACKPROPAGATION THROUGH TIME (BPTT). EEPIS Final Project.

Preview

PDF - Published Version Download (229Kb) | Preview

Abstract

In this final project will be made an application to stock price forecasting using RNN - BPTT. In this final project modeling the input data is the Close Price of shares on the Indonesia Stock Exchange (BEI). Then the data Close Price shares - these shares will be forecasting time series with neural network algorithms recurrent network that BPTT algorithm where the network architecture used is Jordan's RNN. Backpropagation Through Time (BPTT) is a fairly popular training algorithm for recurrent neural network. In recurrent neural network there are several feedback loops in the connection graph. The main concept of the BPTT is to spread the network to the time by putting the same copy of the recurrent neural network and manage network connections back to get the connection between the next set. To produce an accurate forecasting, the parameters in the RNN will be tested, such as learning rate, number of neurons and the number of data. We make this final project is expected to help investors to predict stock price fluctuations so they are able to determine the investment policy of the future with good results. Keywords: Forecasting Time Series, Close Price, Recurrent Neural Network, Backpropagation Through Time

Actions (login required)