/// <summary>
/// Specify a validation set to hold back.
/// </summary>
/// <param name="validationPercent">The percent to use for validation.</param>
/// <param name="shuffle">True to shuffle.</param>
/// <param name="seed">The seed for random generation.</param>
public void HoldBackValidation(double validationPercent, bool shuffle,
int seed)
{
IList <DataDivision> dataDivisionList = new List <DataDivision>();
dataDivisionList.Add(new DataDivision(1.0 - validationPercent)); // Training
dataDivisionList.Add(new DataDivision(validationPercent)); // Validation
_dataset.Divide(dataDivisionList, shuffle,
new MersenneTwisterGenerateRandom((uint)seed));
TrainingDataset = dataDivisionList[0].Dataset;
ValidationDataset = dataDivisionList[1].Dataset;
}
public void Run()
{
var bestMethod = _model.Crossvalidate(5, true) as IMLRegression;
Console.WriteLine($"Training error: {_model.CalculateError(bestMethod, _model.TrainingDataset)}");
Console.WriteLine($"Validation error: {_model.CalculateError(bestMethod, _model.ValidationDataset)}");
var dataDivisions = new List <DataDivision> {
new DataDivision(0.5), new DataDivision(0.5)
};
_dataset.Divide(dataDivisions, true, new MersenneTwisterGenerateRandom());
foreach (var data in dataDivisions[0].Dataset)
{
var result = bestMethod.Compute(data.Input);
Console.WriteLine($"Result: {result}, Actual: {data.Ideal}");
}
EncogDirectoryPersistence.SaveObject(new FileInfo("SvmModel.eg"), bestMethod);
}
