# Modelling volatility for higher frequency data

I'm doing some academic work on volatility forecasting. I've got 1-minute bar data. It is not clear to me what model is best suited for forecasting volatility when higher frequency data is available.

I understand the following families/classes of volatility models exist:

1. (G)ARCH family models
2. Stochastic Volatility
3. Implied Volatility (not applicable because I don't have options prices)
4. Realised volatility

I was wondering, considering I have high frequency data, realised volatility should provide a reasonable approximation. I could potentially calculate the volatility using realised volatility and then use standard time series forecasting methods to forecast this (realised volatility) series?

• I can't answer the question but there are definitely papers by barndorf nielsen ( and other co-authors ) that deal with realized volatility. They probably gives arguments in those as to why forecasting volatility should be done using the realized volatility approach. That should help some. – mark leeds Jul 18 '20 at 22:17