A Multi Headed Artificial Intelligence Approach for Stock Market Trading

Abstract

Stock market prediction remains a challenging task due to market volatility and the complex interplay of multiple factors affecting price movements. While traditional technical analysis and modern machine learning/deep learning approaches have shown promise, they often fall short when used in isolation. This paper presents a novel three-layer approach that combines traditional technical analysis, deep learning, and sentiment analysis for more accurate stock price prediction. The first layer employs technical indicators to capture price trends from historical data. The second layer utilises deep learning models to process comprehensive market data and identify complex patterns. In this second layer, predicted prices are also plotted with historical data, and the buy or sell decision is based on the chart classification. The third layer incorporates real-time sentiment analysis of news and social media to capture market sentiment impact. We evaluate our approach using historical data from major stock exchanges spanning three years (2021-2023). Results demonstrate that our integrated approach significantly outperforms existing methods, achieving lower Mean Absolute Error (MAE) and Mean Squared Error (MSE) scores while maintaining a higher R2. These findings suggest that combining multiple analytical perspectives through a layered architecture can provide more reliable stock market predictions than single-method approaches.