A multivariate LSTM-based deep learning model for stock market prediction

Abstract

Stocks represent ownership in a company and a proportionate claim on its assets and earn-ings. Investors trade stocks via an exchange by buying at a price and selling at a higher price. Due to market volatility forecast it is a necessity for trading to determine the direc-tion of the stock price in order to maximize profit and minimize loss. Traditional methods of stock price predictions include technical and fundamental analysis. The technical deals with historical price movement while fundamental analysis uses the relationship between financial information about the company. However, these predictions methods sometimes fail to yield desired result sometimes due to the influence of factors such as national poli-cies, global and regional economics, psychological, human among many. This work propos-es a prediction model for stock market using LSTM algorithm. Multivariate time series stock price data is obtained from Nigerian Stock Exchange Index to implement the model. The experimental result of the technique is measured using MAPE, MAE, MSE and rRMSE performance metrics. The accuracy of the result shows that the proposed system outper-forms existing traditional and deep learning methods.