From Data to Decisions: Evaluating Deep Learning Models for Stock Price Prediction in Volatile Markets

Predicting Prices in Financial markets, a pivotal challenge and task to overcome due to the volatility of market and nonlinear dynamics. This research studies and evaluates various deep learning models such as, ANN, DQN, LSTM, RNN, XGBoost, CNN and KNN for estimating and predicting price of Tata Steel Stock (NSE: TATASTEEL), across varying timeframes (10 years, 5 years, 1 year, 6 months). Evaluated models were compared using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and computational runtime. Our results show that ANN, DQN and LSTM outperforms others in long-term forecasts, also LSTM along with RNN provide better accuracy for short term time frames. No single model performed finest across varying timeframes, suggesting prospective demand hybrid models. Our finding highlight, trade-offs between accuracy and computational efficiency, directing researchers for model selection in different market condition.