jpmml-sparkml-lightgbm
jpmml-sparkml-lightgbm copied to clipboard
JPMML-SparkML plugin for converting LightGBM-Spark models to PMML
My dataset contains some string features, and I used StringIndexer + OneHotEncoder to encode them. When I put the StringIndexers, OneHotEncoders, VectorAssembler, and LightGBM in a pipeline, and fit the...
Hi @terrytangyuan, I'm currently trying to upgrade the package to use mmlspark 0.16 and also make it is accessible from https://github.com/jpmml/pyspark2pmml. Do you currently use this within java or have...
Current in `BoosterUtils.encodeBinaryClassificationBooster()`, some parameters we used for `MiningModelUtil.createBinaryLogisticClassification()` and `ModelUtil.createPredictedOutput()` are hard-coded. It would be better if we can look into JPMML docs further to see if some of...
Make this into a Spark package and publish at spark-packages.org so it's easy to use from any Spark installation. Instructions: *Third-party packages for Apache Spark should take care to respect...
We only tested the converter internally through our integration. It would be good to add some test cases here similar to what's in [jpmml-sparkml-xgboost](https://github.com/jpmml/jpmml-sparkml-xgboost).
It would be good to add some examples in README.md similar to what's in [jpmml-sparkml-xgboost](https://github.com/jpmml/jpmml-sparkml-xgboost).
Currently we only support conversion from binary classifier. It would be good to support the other types of models: * Regression * Multi-class classification (blocked by https://github.com/Azure/mmlspark/issues/342)