In emerging economies such as Nigeria, predicting stock market volatility is vital for making sound investment and policy decisions. Sudden price swings and shifting market conditions make this task especially challenging. Although established econometric models like ARCH and GARCH are widely used, they often fall short when it comes to capturing nonlinear behaviors and abrupt changes in volatility regimes. To address this gap, we present a hybrid expert system that integrates GARCH-type models, Markov Regime Switching, and Deep Convolutional Neural Networks (CNNs). In our approach, daily stock returns from the Nigerian Stock Exchange between 2012 and 2023 are transformed into Gramian Angular Field (GAF) images, enabling the CNN to detect complex temporal spatial structures. The models are evaluated using RMSE and MAE, and their performance is compared to that of Individual methods. The hybrid system consistently produces more accurate and resilient forecasts, particularly during regime shifts. These findings highlight its value as a decision-support tool for investors, policymakers, and portfolio managers. We recommend the adoption of similar integrated approaches for analyzing volatile markets, particularly in environments with limited or unstable data.