I am trying to estimate an extreme gradiant boosting model using (xgboost) package.
Before writing this question, I looked for infos both on stackoverflow and elsewhere, but the informations that I found did not solve my problem.
Those are the posts that are similar to this that I found, but they have not been solved in a clear way:
human readable rules from xgboost in R
Why is xgboost not plotting my trees?
My problem is with the function xgb.plot.tree, I built my extreme gradiant boosting model with xgboost() function, I created matrix that shows importance of features in a model with xgb.importance(), but when I try to plot a boosted tree model, the function crashes.
Below my R-code:
model_xgb <- xgboost(data=dtrain$data,
label=dtrain$label,
booster="gbtree",
max_depth=6,
eta=0.3,
nthread=2,
nrounds = 150,
verbose=0,
eval_metrics=list("error"),
objective="binary:logistic")
# Calculate feature importance
imp <- xgb.importance(feature_names=dtrain$data@Dimnames[[2]],
model=model_xgb)
print(imp)
# Display tree number 1
xgb.plot.tree(feature_names=dtrain$data@Dimnames[[2]],
model=model_xgb,
n_first_tree=1)
According with the answer of @ArriveW I added the function xgb.dump() and I modified my code as below:
dump_xgb <- xgb.dump(model = model_xgb)
xgb.plot.tree(feature_names=dtrain$data@Dimnames[[2]],
model=model_xgb,
n_first_tree=5)
But despite this, I do not get a chart and this is the result
On github you can find the reproducible example:
Any suggestion please?
