how to predict votility effectly?

[Random_do]

Today is also sort of the end of the options market arbitrage research, since the last time I posted a switch to backtesting with buy-one-sell-one intraday data, and not surprisingly, at the end of the day, either there is no chance, or can't make up for the commissions and slippage.
Tomorrow I'm going to switch to volatility research, and the only major types of forecasting methods available are time series models, machine learning algorithms, and volatility models.
Timing model can only be directly located in the Gage to try machine learning this side have seen with lightqbm to do, I do not know the comment area has a big brother has done related to the volatility model this side is slightly more complex, see the paper used in the unlearned hh have to study a little!
So I think I'll start with a machine learning model, can any of you give me some advice?
The purpose of designing a volatility strategy is to design a predictive volatility

 

[lourano.x]

I'd say that the deeper you delve into market microstructure, the more accurate and robust your predictions are likely to be. Just have a look how much effort HFT boutiques push to predict the realised volatility with ML.