Stock Sense: A Hybrid CNN-RL Approach for Financial Market Price Prediction Using News and Market Data

Financial market prediction presents significant difficulties brought on by market data's non-linearity and volatility. Traditional models like RCNN and RNN rely on static decision-making, which limits their ability to adapt to real-time fluctuations. This study suggests a novel hybrid model that combines convolutional neural networks (CNN) for feature extraction with reinforcement learning (RL) for dynamic decision optimization. By processing multimodal data such as market data, news sentiment extracted using a BERT-based model, and technical indicators, the system improves prediction accuracy. The model can now concentrate on important data points thanks to the addition of an attention mechanism, enhancing its ability to predict market trends. The RL component utilizes a dynamic fusion technique that adapts to real-time data, overcoming the limitations of static decision-making. This hybrid approach offers a more reliable, accurate, and adaptive system for financial market prediction, increasing accuracy by up to 60% compared to traditional models. For traders and investors navigating intricate market situations, the suggested framework offers a strong tool for decision-support.