Forecating Composite Stock Price Index (CSPI) Using Long Short Term Memory (LSTM)

Abstract

The Composite Stock Price Index (CSPI) is an index that displays developments the whole movement of the company's share price in the stock market which refers to the Indonesia Stock Exchange (IDX). Before considering investment, investors can predict the Indonesian stock market is up and down by CSPI analysis. The main objective of this research is to propose forecasting model of CSPI using Long Short Term Memory (LSTM). The performance of LSTM model measured by Root Mean Square Error (RMSE). The results showed that the best LSTM models is model with number of neuron in hidden layer and epoch (iterations) were 10 and 10, respectively. The RMSE values achieved from the LSTM models for testing data is 0,0633. Visually, the prediction graph is almost similar with original data.