A simple wrapper around the Microsoft LightGBM library. Documentation for the various library parameters (see LightGBM.Parameters) can be found here.

N.B. The lightgbm executable must be on the system PATH.

N.B. This is alpha-level software and should not be used in production since the API may still change substantially.

The basic usage of the library looks something like this:

 {-# LANGUAGE TemplateHaskell #-}

 [...]

 import           LightGBM ( toCSV
                           , readCsvFile
                           , HasHeader(..)
                           , trainNewModel)
 import qualified LightGBM.Parameters as P
 import           Refined (refineTH)

 let modelFile = "/path/to/model/output"
     trainingData = readCsvFile (HasHeader False) "/path/to/training/data"
     validationData = readCsvFile (HasHeader False) "/path/to/validation/data"
     trainingParams = [ P.App P.Binary
                      , P.Metric [P.BinaryLogloss, P.AUC]
                      , P.TrainingMetric True
                      , P.LearningRate 0.1
                      , P.NumLeaves 63
                      , P.FeatureFraction $$(refineTH 0.8)
                      , P.BaggingFreq $$(refineTH 5)
                      , P.BaggingFraction $$(refineTH 0.8)
                      , P.MinDataInLeaf 50
                      , P.MinSumHessianInLeaf 5.0
                      , P.IsSparse True
                      ]

 modelOut <- trainNewModel trainingParams trainingData validationData
 case modelOut of
     Left err -> ... -- handle the errors
     Right model -> do
         let newData = readCsvFile (HasHeader False) "/path/to/inputs_for_prediction"
             outputFile = "/path/to/prediction_outputs"
         predOut <- predict model [] newData
         case predOut of
             Left err -> ... -- handle the errors
             Right preds -> toCSV outputFile preds