Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Feature selection in your ML model

I have built an ML model (XG boost) to trade the cryptocurrency market and am kind of unsure on my method for including features, there has been times were i add features which increase my objective function (im using negative log loss) but also increase my backtested returns. Also there has been occasions where i remove features who have 0 explanatory power to the model but when they are removed my returns decrease while my objective function also decreases. Should I just being pay attention to my objective function as I was thinking I am possibly curve fitting if i make decisions based off my backtest's profitability