A Review of Methods for Predicting Stock Volatility: A Comparison of Traditional Econometric Models, Implied Volatility, and Machine Learning Methods

Stock volatility forecasting serves as a crucial underpinning for derivatives pricing, asset allocation, and risk management. This paper examines volatility proxy variables, empirical applications, as well as forecasting evaluation metrics, and then reviews representative models and empirical evidence across three research pathways. The comparison concerns the information basis and forecasting horizon, taking into account explanatory power, stability, and economic value. A comprehensive review of the literature suggests that traditional models remain a robust benchmark in scenarios with limited data, in which emphasis is placed on interpretability and reproducibility. Implied volatility can provide incremental information beyond historical data; however, its predictive power is influenced by variance risk premiums, liquidity, and maturity matching. Machine learning offers greater flexibility in high-dimensional data environments, but its advantages depend on rigorous time-series validation and control of information leakage. Because of inconsistencies in proxy variables, loss functions, sample partitioning, and market conditions, existing evidence fails to support the universal superiority of any single method across all tasks. Future analysis should conduct cross-market tests under a unified evaluation framework and integrate statistical forecasting accuracy with economic value in risk management and investment decision-making.