public static void CreateModel()
{
MLContext mlContext = new MLContext();
IDataView trainingDataView = mlContext.Data.LoadFromTextFile <ModelInput>(
path: TrainDataFilePath,
hasHeader: true,
separatorChar: ',',
allowQuoting: true,
allowSparse: false);
var dataProcessPipeline = mlContext.Transforms.Categorical.OneHotEncoding(
new[]
{
new InputOutputColumnPair("District", "District"),
new InputOutputColumnPair("Type", "Type"),
new InputOutputColumnPair("BuildingType", "BuildingType")
})
.Append(mlContext.Transforms.Concatenate("Features", new[] { "District", "Type", "BuildingType", "Size", "Floor", "TotalFloors", "Year" }));
FastTreeTweedieTrainer trainer = mlContext.Regression.Trainers.FastTreeTweedie(new FastTreeTweedieTrainer.Options()
{
NumberOfLeaves = 20, MinimumExampleCountPerLeaf = 10, NumberOfTrees = 500, LearningRate = 0.07684207f, Shrinkage = 1.057825f, LabelColumnName = "Price", FeatureColumnName = "Features"
});
EstimatorChain <RegressionPredictionTransformer <FastTreeTweedieModelParameters> > trainingPipeline = dataProcessPipeline.Append(trainer);
ITransformer model = trainingPipeline.Fit(trainingDataView);
mlContext.Model.Save(model, trainingDataView.Schema, ModelFilePath);
}
public void TweedieRegressorEstimator()
{
// Pipeline.
var pipeline = new FastTreeTweedieTrainer(Env, "Label", "Features", advancedSettings: s => {
s.EntropyCoefficient = 0.3;
s.OptimizationAlgorithm = BoostedTreeArgs.OptimizationAlgorithmType.AcceleratedGradientDescent;
});
TestEstimatorCore(pipeline, GetRegressionPipeline());
Done();
}
public void TweedieRegressorEstimator()
{
var dataView = GetRegressionPipeline();
var trainer = new FastTreeTweedieTrainer(Env, "Label", "Features", advancedSettings: s =>
{
s.EntropyCoefficient = 0.3;
s.OptimizationAlgorithm = BoostedTreeArgs.OptimizationAlgorithmType.AcceleratedGradientDescent;
});
TestEstimatorCore(trainer, dataView);
var model = trainer.Train(dataView, dataView);
Done();
}
