Financial institutions need accurate risk management to stay competitive and keep the economy stable. A new method called SVR-GARCH-KDE hybrid is proposed to forecast Value-at-Risk (VaR) more effectively. This method combines support vector regression (SVR) for mean and volatility modeling, generalized autoregressive conditional heteroscedasticity (GARCH) for volatility, and kernel density estimation (KDE) for VaR calculation. The hybrid model outperforms traditional models in forecasting one-day-ahead VaR for different financial indices, especially when compared to models using a normal distribution.