如何更改XGboost中的损耗矩阵（在R中）？

Question

I want to classify a binary variable with cost of a false positive higher than a false negative.

In rpart package we use the loss matrix , adding parms = list(loss=matrix(c(0,1,5,0),nrow=2)) for the cost of mis-classifying a negative example as positive is 5 times higher than the cost of mis-classifying a positive example as negative.

How can I do that with XGboost ?

Answer 1

Are you looking for scale_pos_weight parameter?

https://github.com/dmlc/xgboost/blob/master/doc/parameter.md

scale_pos_weight, [default=1] Control the balance of positive and negative weights, useful for unbalanced classes. A typical value to consider: sum(negative cases) / sum(positive cases) See Parameters Tuning for more discussion. Also see Higgs Kaggle competition demo for examples: R, py1, py2, py3

You can use it something like:

clf = xgb.XGBRegressor(objective='binary:logistic', 
                       scale_pos_weight= 5,
                       max_depth=3,
                       n_estimators=100)

in python, sklearn api.

Answer 2

Assuming you are using xgboost package You can use watchlist parameter. It is a list of xgb.DMatrix , each of them tagged with a name. You can use eval.metric parameter, multiple evaluation metrices are also allowed.

watchlist <- list(train=dtrain, test=dtest)

bst <- xgb.train(data=dtrain, max.depth=2, eta=1, nthread = 2,
eval.metric = "error", eval.metric = "logloss", nround=2,
watchlist=watchlist, objective = "binary:logistic")

If the extensive list of metrices in xgboost github pages does not suffice your need then as they say you can generate you own metrices, eg a weighted sum of false positive and false negative where false positive is weighted five times more than the false negative.