In emerging stock exchange markets, the ability to predict stock price volatility is critical for managing financial risk. This paper presents a novel approach to forecasting volatility by employing a hybrid machine learning model. We introduce a two-tiered hybrid model which integrates the robust predictive power of eXtreme Gradient Boosting (XGBoost) with a corrective step using linear regression to refine predictions further. This model is designed to address the challenges posed by the inherent volatility in emerging markets. Our methodology leverages historical stock data of the top 30 performing stocks in the Egyptian stock exchange listed under the EGX30 index to predict volatility, employing a machine learning framework that surpasses traditional methods in terms of accuracy. The results demonstrate that our model achieves a mean absolute percentage error (MAPE) of less than 10% for 20 out of 30 stocks analyzed, with an overall MAPE of 9.35%. Additionally, the model consistently delivers a mean directional accuracy of 82.1% across the index constituents. These outcomes not only provide a significant improvement over existing benchmarks but also contribute to the literature by offering a refined approach to managing the unpredictability of volatile markets through innovative machine learning techniques.