ModelOriented / forester

Trees are all you need
https://modeloriented.github.io/forester/
GNU General Public License v3.0
108 stars 14 forks source link

lightgbm params #95

Closed inoddy closed 1 year ago

inoddy commented 1 year ago

library(forester) data("titanic", package = 'DALEX') train_output <- train(data = titanic, "survived", type = "classification", engine = c("ranger", "xgboost", "decision_tree", "lightgbm", "catboost"))

produced after a few minutes

Error in lightgbm::lgb.train(params = params, data = data$lightgbm_data, : object 'params' not found

yet if I omit the type argument it runs ok and does a classification

HubertR21 commented 1 year ago

There was a mistake in the documentation. Instead of 'classification' there should be 'binary_clf' (as there are plans for adding multi-label classification in the future). The fix of that issue will be available in the next package version.