Optimized Bi-GRU model-basedStock Market Prediction: Bigdata Consideration of Stock and News Data

Abstract

Sentiment analysis plays a vital role in making informed decisions about business investments in stock markets. It is crucial for identifying an organization's or company's business through stock analysis. Predicting stock prices can be challenging due to their unstable nature, influenced by various factors such as politics, economics, and leadership changes. Historical or textual data alone may not be enough for efficient prediction. Incorporating news sentiment data with stock price data can significantly improve the accuracy of predictions. To this end, we have developed a prediction framework that utilizes both stock price and news sentiment data. The framework initially retrieves technical indicator-dependent features such as Moving Average Convergence Divergence (MACD), Moving Average (MA), and Relative Strength Index (RSI) from stock data. The news data then undergoes specific processes such as pre-processing, feature extraction, and categorization to identify sentiments. In the final categorization stage, the final prediction takes place by the Optimized Bi-GRU model, where the training is carried out by hybrid optimization model that includes Pelican optimization. We conducted a parametric and non-parametric analysis of our proposed POA by altering the parameters.