Hybrid Deep Learning Framework for Multiscale Forecasting of Stock Price Dynamics in Nigerian Banking Stocks

Forecasting stock price movements and volatility is increasingly important in emerging financial markets such as Nigeria, where economic conditions, regulatory changes, and investor sentiment influence market performance. This study examined stock price dynamics and volatility in five selected Nigerian banks using a hybrid Recurrent Neural Network-Convolutional Neural Network–Long Short-Term Memory (RNN–CNN–LSTM) architecture. Daily stock price data for Fidelity Bank Plc, Stanbic IBTC Holdings Plc, United Bank for Africa (UBA) Plc, Wema Bank Plc, and Zenith Bank Plc were obtained from the Nigerian Exchange (NGX) for the period from 3 January 2014 to 31 December 2024. The hybrid model integrated RNN, CNN, and LSTM techniques to capture temporal dependencies, local patterns, and long-term relationships in stock-price movements. The model demonstrated strong predictive performance, achieving an RMSE of 0.1880 for Fidelity Bank and 0.4791 for UBA, with corresponding MSE and MAE values of 0.0353 and 0.1562 for Fidelity, and 0.2295 and 0.3911 for UBA compared to the standalone models. The findings identified three major phases: relative stability before 2020, heightened volatility during the COVID-19 period (2020–2021), and sustained recovery from 2022 onwards, particularly for Zenith Bank, UBA, and Fidelity Bank. Evidence of volatility clustering was also observed, indicating persistence in periods of high and low volatility. Overall, the study demonstrates the potential of the hybrid RNN–CNN–LSTM framework for forecasting stock price dynamics and volatility in Nigerian banking stocks, providing useful insights for investors, financial institutions, and policymakers.