Published Paper Details:

The inherent volatility of financial markets presents a significant challenge to accurate price forecasting. This paper introduces Algorion, a novel deep learning framework for stock price prediction. Algorion utilizes a Long Short-Term Memory (LSTM) network trained on an extensive set of historical, technical, and fundamental data to forecast future price movements. To evaluate the model's effectiveness, we conduct a rigorous backtesting experiment on a diverse portfolio of stocks from the NIFTY 50 index over a ten-year period. The results demonstrate that the Algorion framework achieves superior prediction accuracy (measured by RMSE and MAE). Furthermore, when its predictions are used to generate systematic trading signals, the resulting strategy yields significantly higher risk-adjusted returns (measured by Sharpe and Sortino ratios) compared to traditional baseline strategies. An ablation study further quantifies the value-add of integrating a rich feature set, which improves performance over a basic LSTM model. Our findings establish the efficacy of sophisticated deep learning systems in creating more robust and accurate stock price prediction models.