An LSTM-Based Approach to Predict Stock Price Movement for IT Sector Companies

• Published in: International journal of cognitive informatics & natural intelligence 2021-10, Vol.15 (4), p.1-12
• Main Authors: Rammurthy, Shruthi Komarla, Patil, Sagar B
• Format: Article
• Language: English
• Subjects: Company structure; Forecasting; Long short-term memory; Mathematical models; Neural networks; Predictions; Prices and rates; Recurrent neural networks; Stock markets; Stocks; Time series
• ISSN: 1557-3958; 1557-3966
• DOI: 10.4018/IJCINI.20211001.oa3

A stock market is an aggregation of buyers and sellers where issuance, buying, and selling of stocks happen. Predicting stock price is a significant concern due to volatility. Historical stock price and historical price data reveal the effect of such factors. Since stock data is time series and prediction can be made accurately with time series forecasting model. LSTM (Long Short Term Memory) model, a particular kind of RNN (Recurrent Neural Network), based on time series forecasting used to predict stock price. LSTM doesn’t have long term dependencies because of its distinctive structure. The study focuses on major IT firms considering the company’s low and high prices. But, mid-price, which is a mean of the low and close price, is considered for the prediction. LSTM based methodology employing mid-price is effective in predicting values compared to other attributes and accuracy of prediction using the LSTM model. We conclude with the present model is more efficient in stock price prediction with a decrease in mean square error.