In trying to build a ML powered trading strategy, one of the most important tasks is to correctly label the data so that the results of whatever classification algo you are using will be properly matched to the actual PnL function of the trading strategy.
Probably the most popular effort in literature is the so-called triple barrier labeling of De Prado.
However, I am wondering if other efforts exist that are already implemented in R/SQL and that specifically take into account:
- transaction cost
- stop loss
- take profit
- market volatility
- other market microstructure facts