Predicting Realized Volatility using XGBoost(I can’t actually predict anything)

I'm trying to perdict if RV(realized volatility) will explode or not ('classify') using an XGB classifier but the problem is i'm not predicting anything i.e my positives are '0' I'm thinking this can be because of poor features,So can anyone help me with this?

I'm using NIFTY Index '5min' OHLC to train my model,the data is from "2025-10-14 11:20:00" – "2026-01-02 15:25:00" (IST),
log_returns are calculated for consecutive periods,
short_term_ema_crossover is difference between ema_9 & ema_21,
long_term_ema_crossover is ema_44 & ema_101,
vrp is rv_5 – vix

vix_momentum is vix difference for consecutive periods

vix_ema_crossover is vix_9 – vix_21
label is the target which is 1, if target_rv_5 >1 & target_rv_10 > 1 (target_rv is 1 if RV for n+1 period > 1.1*(n) period )

(Note:I'm using VIX as proxy for IV as I don't have historical data and i don't know how to set up IV for my dataset TvT)