Real Time Stock Market Prediction using GPU Enabled Hybrid Model

The financial markets are becoming increasingly dynamic, with real-time data playing a crucial role in informed decision-making. Traditional stock price prediction models often rely solely on historical numerical data, which can limit their adaptability in rapidly changing environments. In this paper, we propose a novel approach that leverages GPU acceleration and deep learning to process image-based representations of corporate data and predict stock prices in real time. Our architecture combines Convolutional Neural Networks (CNNs) for feature extraction and Long Short-Term Memory (LSTM) networks for temporal pattern recognition, enabling a more robust and dynamic prediction model. The system is trained using image datasets representing corporate stock trends and further enhanced using real-time stock data fetched via the Yahoo Finance API (y finance). By implementing the solution on Google Colab with CUDA-enabled GPU support, we achieve significant improvements in training speed and model efficiency. The model demonstrates the ability to simulate real-time predictions of corporate stock prices for each month of the year 2025, using companies such as TCS and Infosys as examples. Our results show a promising correlation between predicted and real-time fluctuated stock values, offering a fresh direction for integrating visual data and live market analytics in financial forecasting.