Potentially skim down the momentum variables in the training feature set
At the time of creating this issue, we see some redundancy in the use of momentum-like variables in the feature set, as illustrated by the significant correlations below:
Assess whether the inclusion of say, p_T(B)
and \sum p_T
is necessary, by inspecting the alteration, or lack thereof, of the ROC score whence removing one of the two highly correlated variables.