Application of time series models in stock market volatility forecasting: A review of the literature in the past decade

Accurately forecasting stock market volatility has become crucial, especially as financial markets become more complex and unpredictable. In the past decade, researchers have significantly advanced forecasting methods, moving from basic statistical tools towards sophisticated machine learning and hybrid solutions. Initially, classical models such as GARCH and stochastic volatility proved effective, particularly in capturing asymmetrical patterns, yet these models often fell short in handling complex nonlinear market behaviors. To address this gap, recent approaches involving neural networks and support vector machines (SVMs) have shown notable improvements, largely because of their adaptability to volatile market changes. However, these advanced methods come with their own practical challenges. Hybrid approaches, which combine econometric models with deep learning techniques, have also emerged as powerful tools, although their implementation can be challenging. This review critically evaluates these diverse methodologies, highlighting their practical strengths and limitations. By synthesizing current empirical findings and theoretical insights, the review identifies key challenges, including data scarcity, parameter complexity, and issues related to real-time forecasting and proposes future research directions. Emphasis is placed on integrating multiple data sources and adopting flexible optimization techniques, which could significantly enhance predictive capabilities in dynamic financial environments.