clarify on use SUOD for XGBOD

[jackie930]

Hi,

I see from your paper:

demon- strate SUOD’s effectiveness as an end-to-end framework on more complex combination models like unsupervised LSCP [35] and XG- BOD [34]

Thus I am wondering this seems like in progress? Any suggestions on how I implement this feature?
My initial guess is to compute TODS scores via SUOD, then build the XGBoost model separately combine the score and the original features. Any suggestion is very welcomed!