Stock Market Price Prediction using Ebola Optimization Search Algorithm with Bi-Directional Gated Recurrent Unit

Variations in the stock market are highly influenced by financial, environmental, and political parameters, along with multiple internal and external factors and historical stock data. Although information on the stock market is readily available on news platforms and microblogs, stock price prediction (SPP) is a challenging task due to high volatility that fluctuates with external variations that affect the stock market. In order to achieve accuracy in stock market prediction (SMP), this study proposes a Bi-Directional Gated Recurrent Unit (Bi-Gru) with enhanced prediction accuracy. The Ebola Optimization Search Algorithm (EOSA) is employed for hyperparameter tuning through measuring the classifier parameters for enhanced classification accuracy. The raw data is gathered from the Dow Jones Industrial Average (DJIA) database which is used for the model's performance evaluation. The collected data is then input for preprocessing using a min-max normalization technique to remove noise for enhanced image quality, after which Linear Transform-Based Feature - Mutual Information Cross Feature (LTF-MICF) method is deployed for effective extraction of the relevant features. Then, Bi-Direction Gated Recurrent Unit (Bi-Gru) method is used for classification of emotions based on speech, contributing to improved classification accuracy. The proposed EOSA-Bi-Gru method attains an MAE of 0.253 and an RMSE of 0.451 on the DJIA database, thereby outperforming the existing methods introduced for SPP.