The article introduces a new method for predicting financial risk called SVR-GARCH-KDE hybrid. This method combines different models to better estimate Value-at-Risk (VaR), a key measure in risk management. By using support vector regression (SVR) for volatility and kernel density estimation (KDE) for tail events, the hybrid model can capture complex patterns in financial data. The study shows that the SVR-GARCH-KDE hybrid performs as well as or better than traditional models in forecasting risk for various financial markets.