search

suiji / Arborist

Scalable decision tree training and inference.
Other
82 stars 14 forks source link

Predict() give different values on original data #20

Closed ghost closed 7 years ago

ghost commented 7 years ago

Hi, I used the example in predict() in the Rborist manual, rb <- Rborist( x, y ) predX <- predict( rb, x )

But predX$yPred is very different from rb$validation$yPred

Is this expected? Original randomForest package gave identical predicted value vector on same operation.

suiji commented 7 years ago

In general, they would be expected to differ:

For each row, the "predict" method aggregates the predictions yielded by every tree. The "validate" method, however, aggregates only predictions yielded by those tees for which the given row is not bagged.