International Journal of Drug Delivery Technology
Volume 16, Issue 2, 2026

Hybrid Deep Learning Technique for Stock Price Prediction

Raghav Gupta1, Dr. Nipun Chhabra2

1Research Scholar, Faculty of Computational Sciences, GNA University, Phagwara-144401, Punjab. Contact no: 8360857556. Email: raghav.gupta680@gmail.com. ORCID: 0000-0001-9791-4287

2Associate Professor, School of Engineering, Design & Automation, GNA University, Phagwara-144401, Punjab. Email: nipun.chhabra@gnauniversity.edu.in. ORCID: 0000-0001-5318-5043


ABSTRACT

Background: This study comes up with a hybrid deep learning model to forecast the prices of stocks of Nifty LargeMidcap 250 Index for five years. It makes use of CNN-LSTM hybrid structure. Sentiment analysis combines News on the social media. Wavelet denoising, PCA dimensionality reduction and walk-forward validation are also methodology. Stock Data represents NSE archives 2010-2024.

Results: Results indicate that CNN-LSTM is better in metrics than LSTM and GRU baselines. RMSE is 5.37 in the case of BPCL and 15.58 in the case of TATA Motors in CNN-LSTM. All stocks show the same patterns of low-error in MAE. MAPE shows outstanding 0.03 values of INFY and TATA Motors. The maximum R² is 0.9976 of INFY and 0.995 of CIPLA. The model fills the long-horizon forecasting gaps in the case of blended indices.

Conclusion: Findings justify practical usefulness of CNN-LSTM as it provides the best results.

Keywords: CNN-LSTM, RMSE, Nifty LargeMidcap 250, LSTM, MAPE, LSTM, TATA MOTORS, sentiment, technical indicators, SHAP.

How to cite this article: Gupta R, Chhabra N. Hybrid Deep Learning Technique for Stock Price Prediction. Int J Drug Deliv Technol. 2026;16(2): 662-668. DOI: 10.25258/ijddt.16.2.70

Source of support: Nil.

Conflict of interest: None