Stock Market Prediction Using Long Short-Term Memory (LSTM)

Abstract

The stock market is one of the most unpredictable and highly concerned places in the world. There is no fundamental way to forecast stock market share prices. So people think stock market prediction is a gamble. Nevertheless, it is possible to generate a constructive pattern by using different types of algorithms and predict the share price. But when the characteristics are complex, and the largest portion of these classification methods are linear, resulting bad performance in class label prediction. In this paper we suggest a non-linear technique based on the Long Short-Term Memory (LSTM) architecture. According to studies LSTM-based models predict time and sequential models better than other models and RNN is the first algorithm with an internal memory that remembers its input, making it perfect for sequential data machine learning issues. For our experiment we collected the share market data from a particular company named Beximco for the last 11 years. To reassert the effectiveness of the system different test data are used. This work introduces a robust method that can predict stock price accurately based on LSTM.