The LightGBM model output is quite different to the expected result.

[shipengAlan]

I exported my tree based regression model to PMML format. And based on this library I load the model file, evaluate it with unit cases and found its output score is quite different to the result when training. Do you know the possible reasons? Thanks a lot

[scorebot]

About the results of the model prediction, you can refer to the third part Understand the result values of https://github.com/autodeployai/pmml4s#use-in-scala

There are some models that could define output fields without the predicted value, for example only probabilities of each category, then if you expect all the possible results, you can call the method model.setSupplementOutput(true) to get all results including the final predicted value.

[scorebot]

@shipengAlan Please let me know if you still have a problem