Volatility Forecast

I am trying to use historical data to more accurately predict the future volatility. This will include share movements (high-low-close) and implied volatility.

Hence, I would like a regression to place optimal coefficients on the implied volatility and past daily movements to forecast the next 20 day volatility (eg will put a exponentially decaying on previous daily price range plus some contribution from implied volatility to most accurately predict the next 20 day volatility).

Ordinary least squares regression requirements are often unsatisfied in time series. Consider autogregressive methods, such as ARIMA