A Novel Approach to Stock Price Direction and Price Prediction Based on News Sentiments

Abstract

Forecasting stock trends guide investment management, financial policy, and the country’s economic growth. Investor-generated textual information has impacted stock movements across media channels in recent years. Most sentiment index studies weigh linguistic content equally. Such studies ignore that the sentiment index’s impact on the stock market decreases over time. This study analyses stock indices using dual classifier coupling and sentiment analysis. A dual clas sifier is created by combining two popular classifiers, Decision Tree (DT) with Convolution Bi-Directional Gated Recurrent Unit (GRU). The proposed model is tested using Reliance Industries shares. The adjusted sentiment index improved overall accuracy in the Reliance Industries stock news sentiment analysis case study by 84.12 percent. The investor sentiment indicator improves stock index trend prediction, as shown by a 3.16 RMSE (Root Mean Squared Error) and 0.97 R2(Coefficient of determination) reduction. The adjusted sentiment index improves predicted accuracy considerably. The investors’ sentiments improve the overall results in Reliance Industries’ stock price prediction with our fusion of pro posed VADER (Valence Aware Dictionary and sEntiment Reasoner) and CNN + BDGRU models compared to benchmark models