Convert lookup table into xgboost model

user_9064266 0
Viewed 117

I'm trying to build a binary:logistic xgboost model, but I want the initial estimate of probability to be the value from a lookup table (i.e. a particular set of x values has a single given y value based on the lookup table). I've found the xgb_model argument, but I don't know how to convert a table with probabilities into an xgb_model where the predictor is perfectly preserved.

The challenges here are

  • The lookup table is not appropriate to be used as a binary:logistic model as the predictor values are not limited to 0 and 1
  • I don't know how to ensure that the final output is a 1-1 match. I'm considering something which is significantly overfitted, but I don't know how to implement that in xgboost.
xgboost r
1 Answers

I don't think you need to convert the table with probabilities into an xgb_model here. Also, this argument is not valid "The lookup table is not appropriate to be used as a binary:logistic model as the predictor values are not limited to 0 and 1" because, the predictor values can be numeric but not necessarily binary to feed into xgb_model. Only response[y] has to be binary if you need to build a binary classification model.

Can you please elaborate more about the table that you have mentioned? Please provide a small example for the predictor values[x] and the response[y] and the result you expect as the output.

Thank you

user_73193720
Related