Volatility, which measures price fluctuations, is crucial in finance but not directly observable. A method using maximum likelihood was applied to different stochastic volatility models to predict future returns based on current volatility. The models studied include expOU, OU, and Heston, showing good performance in various market indexes. The method naturally accounts for long-range volatility auto-correlation and return-volatility cross-correlation, despite not being explicitly included in the models.