using System;

using System.IO;

using System.Text;

using System.Text.Json;

using Microsoft.ML;

using Microsoft.ML.Data;

​

public class Program

{

    public static void Main()

    {

        var mlContext = new MLContext();

        var dataset = mlContext.Data.LoadFromTextFile<ModelInput>(WriteDataset(numRows: 100), hasHeader: false);    

        var split = mlContext.Data.TrainTestSplit(dataset, testFraction: 0.3);      

        // Binary PriorPredictor (works)

        Console.WriteLine("Binary PriorPredictor...");

        var baselineBinaryPipeline = mlContext.BinaryClassification.Trainers.Prior(labelColumnName: "Label");

        var binaryModel = baselineBinaryPipeline.Fit(split.TrainSet);

        var binaryPredictions = binaryModel.Transform(split.TestSet);

        var baselineBinaryMetrics = mlContext.BinaryClassification.Evaluate(binaryPredictions, "Label", "Score");

        PrintMetrics(baselineBinaryMetrics);

        // Multiclass OVA-AveragedPerceptron (works)

        Console.WriteLine("\nMulticlass OVA-AveragedPerceptron...");

        var ovaAveragedPerceptronPipeline = mlContext.Transforms.Conversion.MapValueToKey("Label", "Label")

            .Append(mlContext.MulticlassClassification.Trainers.OneVersusAll(mlContext.BinaryClassification.Trainers.AveragedPerceptron(labelColumnName: "Label", featureColumnName: "Features"), labelColumnName: "Label"));

        var ovaAveragedPerceptronModel = ovaAveragedPerceptronPipeline.Fit(split.TrainSet);

        var ovaAveragedPerceptronPredictions = ovaAveragedPerceptronModel.Transform(split.TestSet);

        var ovaAveragedPerceptronMulticlassMetrics = mlContext.MulticlassClassification.Evaluate(ovaAveragedPerceptronPredictions, "Label", "Score");

        PrintMetrics(ovaAveragedPerceptronMulticlassMetrics);

        // Multiclass OVA-PriorPredictor (fails)

        Console.WriteLine("\nMulticlass OVA-PriorPredictor...");

        var baselineMulticlassPipeline = mlContext.Transforms.Conversion.MapValueToKey("Label", "Label")

            .Append(mlContext.MulticlassClassification.Trainers.OneVersusAll(mlContext.BinaryClassification.Trainers.Prior(labelColumnName: "Label"), labelColumnName: "Label"));

        var multiclassModel = baselineMulticlassPipeline.Fit(split.TrainSet);

        var multiclassPredictions = multiclassModel.Transform(split.TestSet);

        var baselineMulticlassMetrics = mlContext.MulticlassClassification.Evaluate(multiclassPredictions, "Label", "Score");

        PrintMetrics(baselineMulticlassMetrics);

    }

​

    private static void PrintMetrics<T>(T metrics) {

        var options = new JsonSerializerOptions(){ WriteIndented = false };

        var jsonString = JsonSerializer.Serialize(metrics, options);

        Console.WriteLine(jsonString);

    }

    private static string WriteDataset(long numRows)

    {

        var random = new Random();

        var dataDirectoryName = "DataDir";

        Directory.CreateDirectory(dataDirectoryName);

        var fileName = Path.Combine(dataDirectoryName, $"Data.csv");

​

        using (var fs = File.CreateText(fileName))

        {

            for (int line = 0; line < numRows; line++)

            {

                var sb = new StringBuilder();

                sb.Append(random.NextDouble() > 0.5 ? "true" : "false");

                for (int i = 1; i <= 10; i++)

                    sb.Append($"\t{random.NextDouble()}");

                fs.WriteLine(sb.ToString());

            }

        }

​

        return fileName;

    }

    public class ModelInput

    {

        [ColumnName("Label"), LoadColumn(0)]

        public bool Label { get; set; }

​

        [ColumnName("Features"), LoadColumn(1, 10)]

        public float Features { get; set; }

    }

}