Model options for numeric and categorical

Specific Domains Machine Learning
1

Hi All,

I am wondering if there is something similar in Julia to Python’s Catboost, i.e. a gradient boosted determinant model that allows for categorical and numeric data?

I found a Catboost.jl package, but it is reliant on mini-conda and my virtual environment won’t allow for this - and I would like a 100% Julia solution. I was also interested in any learning to rank (LTR) models such as LightGBM ranker - julia has a LightGBM but they haven’t imported the ranking feature for some reason?

The one package that got close to this is JLBoost.jl but is deprecated :confused:

I would be open to hearing about other packages / models that could achieve these as well.

2

EvoTrees.jl is an actively maintained gradient tree boosting package that is 100% Julia. I’m not sure if there is support for categorical variables, however. If you don’t mind one hot encoding these, you can combine EvoTrees models with OneHotEncoder or ContinuousEncoder in an MLJ pipeline, as EvoTrees.jl models have an MLJ interface.

If you have experience in machine learning and are coming from another platform, you man find MLJ for Data Scientists in Two Hours a good starting point. Disclaimer: I am the lead developer of MLJ.

There is also LightGBM.jl and XGBoost.jl, which are C-wrappers.