[C#/ML.NET] 다중 클래스 분류 모델 사용하기

TestProject.zip

다운로드

▶ IssueData.cs

using Microsoft.ML.Data;

namespace TestProject
{
    /// <summary>
    /// 이슈 데이터
    /// </summary>
    public class IssueData
    {
        //////////////////////////////////////////////////////////////////////////////////////////////////// Property
        ////////////////////////////////////////////////////////////////////////////////////////// Public

        #region ID - ID

        /// <summary>
        /// ID
        /// </summary>
        [LoadColumn(0)]
        public string ID { get; set; }

        #endregion
        #region 영역 - Area

        /// <summary>
        /// 영역
        /// </summary>
        [LoadColumn(1)]
        public string Area { get; set; }

        #endregion
        #region 제목 - Title

        /// <summary>
        /// 제목
        /// </summary>
        [LoadColumn(2)]
        public string Title { get; set; }

        #endregion
        #region 설명 - Description

        /// <summary>
        /// 설명
        /// </summary>
        [LoadColumn(3)]
        public string Description { get; set; }

        #endregion
    }
}

 

728x90

 

▶ IssuePrediction.cs

using Microsoft.ML.Data;

namespace TestProject
{
    /// <summary>
    /// 이슈 예측
    /// </summary>
    public class IssuePrediction
    {
        //////////////////////////////////////////////////////////////////////////////////////////////////// Property
        ////////////////////////////////////////////////////////////////////////////////////////// Public

        #region 영역 - Area

        /// <summary>
        /// 영역
        /// </summary>
        [ColumnName("PredictedLabel")]
        public string Area;

        #endregion
    }
}

 

300x250

 

▶ Program.cs

using System;
using System.IO;

using Microsoft.ML;
using Microsoft.ML.Data;

namespace TestProject
{
    /// <summary>
    /// 프로그램
    /// </summary>
    class Program
    {
        //////////////////////////////////////////////////////////////////////////////////////////////////// Field
        ////////////////////////////////////////////////////////////////////////////////////////// Static
        //////////////////////////////////////////////////////////////////////////////// Private

        #region Field

        /// <summary>
        /// 애플리케이션 경로
        /// </summary>
        private static string _applicationPath => Path.GetDirectoryName(Environment.GetCommandLineArgs()[0]);

        /// <summary>
        /// 훈련 데이터 파일 경로
        /// </summary>
        private static string _trainingDataFilePath => Path.Combine(_applicationPath, "Data", "issues_train.tsv");

        /// <summary>
        /// 테스트 데이터 파일 경로
        /// </summary>
        private static string _testDataFilePath => Path.Combine(_applicationPath, "Data", "issues_test.tsv");

        /// <summary>
        /// 모델 파일 경로
        /// </summary>
        private static string _modelFilePath => Path.Combine(_applicationPath, "Data", "model.zip");

        /// <summary>
        /// 모델 컨텍스트
        /// </summary>
        private static MLContext _context;

        /// <summary>
        /// 예측 엔진
        /// </summary>
        private static PredictionEngine<IssueData, IssuePrediction> _predictionEngine;

        /// <summary>
        /// 모델
        /// </summary>
        private static ITransformer _model;

        /// <summary>
        /// 훈련 데이터 뷰
        /// </summary>
        private static IDataView _trainingDataView;

        #endregion

        //////////////////////////////////////////////////////////////////////////////////////////////////// Method
        ////////////////////////////////////////////////////////////////////////////////////////// Static
        //////////////////////////////////////////////////////////////////////////////// Private

        #region 프로그램 시작하기 - Main()

        /// <summary>
        /// 프로그램 시작하기
        /// </summary>
        private static void Main()
        {
            Console.WriteLine("BEGIN MAIN FUNCTION");

            _context = new MLContext(seed: 0);

            Console.WriteLine("BEGIN SET TRAINING DATA VIEW");

            _trainingDataView = _context.Data.LoadFromTextFile<IssueData>(_trainingDataFilePath, hasHeader: true);

            Console.WriteLine("END SET TRAINING DATA VIEW");

            Console.WriteLine("BEGIN SET PIPE LINE");

            var pipeline = GetPipeLine();

            Console.WriteLine("END SET PIPE LINE");

            _model = GetModel(_trainingDataView, pipeline);

            Console.WriteLine("BEGIN PREDICT SINGLE ITEM");

            _predictionEngine = _context.Model.CreatePredictionEngine<IssueData, IssuePrediction>(_model);

            IssueData issueData = new IssueData()
            {
                Title       = "WebSockets communication is slow in my machine",
                Description = "The WebSockets communication used under the covers by SignalR looks like is going slow in my development machine.."
            };

            IssuePrediction issuePrediction = _predictionEngine.Predict(issueData);

            Console.WriteLine("--------------------------------------------------");
            Console.WriteLine($"AREA : {issuePrediction.Area}");
            Console.WriteLine("--------------------------------------------------");

            Evaluate(_trainingDataView.Schema);

            SaveModel(_context, _trainingDataView.Schema, _model);

            PredictIssue();

            Console.WriteLine("END MAIN FUNCTION");
        }

        #endregion

        #region 파이프 라인 구하기 - GetPipeLine()

        /// <summary>
        /// 파이프 라인 구하기
        /// </summary>
        /// <returns>파이프 라인</returns>
        private static IEstimator<ITransformer> GetPipeLine()
        {
            Console.WriteLine("BEGIN GET PIPE LINE FUNCTION");

            var pipeline = _context.Transforms.Conversion.MapValueToKey(inputColumnName : "Area", outputColumnName : "Label")
                .Append(_context.Transforms.Text.FeaturizeText(inputColumnName : "Title"      , outputColumnName : "TitleFeaturized"      ))
                .Append(_context.Transforms.Text.FeaturizeText(inputColumnName : "Description", outputColumnName : "DescriptionFeaturized"))
                .Append(_context.Transforms.Concatenate("Features", "TitleFeaturized", "DescriptionFeaturized"))
                .AppendCacheCheckpoint(_context)
                .Append(_context.MulticlassClassification.Trainers.SdcaMaximumEntropy("Label", "Features"))
                .Append(_context.Transforms.Conversion.MapKeyToValue("PredictedLabel"));

            Console.WriteLine("END GET PIPE LINE FUNCTION");

            return pipeline;
        }

        #endregion
        #region 모델 구하기 - GetModel(trainingDataView, pipeline)

        /// <summary>
        /// 모델 구하기
        /// </summary>
        /// <param name="trainingDataView">훈련 데이터 뷰</param>
        /// <param name="pipeline">파이프 라인</param>
        /// <returns>모델</returns>
        private static ITransformer GetModel(IDataView trainingDataView, IEstimator<ITransformer> pipeline)
        {
            Console.WriteLine("BEGIN GET MODEL FUNCTION");

            var model = pipeline.Fit(trainingDataView);

            Console.WriteLine("END GET MODEL FUNCTION");

            return model;
        }

        #endregion
        #region 평가하기 - Evaluate(trainingDataViewSchema)

        /// <summary>
        /// 평가하기
        /// </summary>
        /// <param name="trainingDataViewSchema">훈련 데이터 뷰 스키마</param>
        private static void Evaluate(DataViewSchema trainingDataViewSchema)
        {
            Console.WriteLine("BEGIN EVALUATE FUNCTION");

            IDataView testDataView = _context.Data.LoadFromTextFile<IssueData>(_testDataFilePath, hasHeader : true);

            MulticlassClassificationMetrics metrics = _context.MulticlassClassification.Evaluate(_model.Transform(testDataView));

            Console.WriteLine("--------------------------------------------------"      );
            Console.WriteLine("METRICS FOR MULTI-CLASS CLASSIFICATION MODEL - TEST DATA");
            Console.WriteLine("--------------------------------------------------"      );
            Console.WriteLine($"MICRO ACCURACY     : {metrics.MicroAccuracy:0.###}"     );
            Console.WriteLine($"MACRO ACCURACY     : {metrics.MacroAccuracy:0.###}"     );
            Console.WriteLine($"LOG LOSS           : {metrics.LogLoss:#.###}"           );
            Console.WriteLine($"LOG LOSS REDUCTION : {metrics.LogLossReduction:#.###}"  );
            Console.WriteLine("--------------------------------------------------"      );

            Console.WriteLine("END EVALUATE FUNCTION");
        }

        #endregion
        #region 모델 저장하기 - SaveModel(context, trainingDataViewSchema, model)

        /// <summary>
        /// 모델 저장하기
        /// </summary>
        /// <param name="context">ML 컨텍스트</param>
        /// <param name="trainingDataViewSchema">훈련 데이터 뷰 스키마</param>
        /// <param name="model">모델</param>
        private static void SaveModel(MLContext context, DataViewSchema trainingDataViewSchema, ITransformer model)
        {
            Console.WriteLine("BEGIN SAVE MODEL FUNCTION");

            context.Model.Save(model, trainingDataViewSchema, _modelFilePath);

            Console.WriteLine("END SAVE MODEL FUNCTION");
        }

        #endregion
        #region 이슈 예측하기 - PredictIssue()

        /// <summary>
        /// 이슈 예측하기
        /// </summary>
        private static void PredictIssue()
        {
            Console.WriteLine("BEGIN PREDICT ISSUE FUNCTION");

            ITransformer model = _context.Model.Load(_modelFilePath, out var inputSchema);

            IssueData issueData = new IssueData()
            {
                Title       = "Entity Framework crashes",
                Description = "When connecting to the database, EF is crashing"
            };

            _predictionEngine = _context.Model.CreatePredictionEngine<IssueData, IssuePrediction>(model);

            IssuePrediction issuePrediction = _predictionEngine.Predict(issueData);

            Console.WriteLine("--------------------------------------------------");
            Console.WriteLine($"AREA : {issuePrediction.Area}");
            Console.WriteLine("--------------------------------------------------");

            Console.WriteLine("END PREDICT ISSUE FUNCTION");
        }

        #endregion
    }
}