[C#/ML.NET] 이진 클래스 분류 모델 사용하기

TestProject.zip

다운로드

▶ SentimentData.cs

using Microsoft.ML.Data;

namespace TestProject
{
    /// <summary>
    /// 감정 데이터
    /// </summary>
    public class SentimentData
    {
        //////////////////////////////////////////////////////////////////////////////////////////////////// Field
        ////////////////////////////////////////////////////////////////////////////////////////// Public

        #region Field

        /// <summary>
        /// 감정 텍스트
        /// </summary>
        [LoadColumn(0)]
        public string SentimentText;

        /// <summary>
        /// 감정
        /// </summary>
        [LoadColumn(1)]
        [ColumnName("Label")]
        public bool Sentiment;

        #endregion
    }
}

 

728x90

 

▶ SentimentPrediction.cs

using Microsoft.ML.Data;

namespace TestProject
{
    /// <summary>
    /// 감정 예측
    /// </summary>
    public class SentimentPrediction : SentimentData
    {
        //////////////////////////////////////////////////////////////////////////////////////////////////// Property
        ////////////////////////////////////////////////////////////////////////////////////////// Public

        #region 예측 - Prediction

        /// <summary>
        /// 예측
        /// </summary>
        [ColumnName("PredictedLabel")]
        public bool Prediction { get; set; }

        #endregion
        #region 확률 - Probability

        /// <summary>
        /// 확률
        /// </summary>
        public float Probability { get; set; }

        #endregion
        #region 점수 - Score

        /// <summary>
        /// 점수
        /// </summary>
        public float Score { get; set; }

        #endregion
    }
}

 

300x250

 

▶ Program.cs

using System;
using System.Collections.Generic;
using System.IO;

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

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

        #region Field

        /// <summary>
        /// 레이블 파일 경로
        /// </summary>
        private static readonly string _labelFilePath = Path.Combine(Environment.CurrentDirectory, "Data", "label.txt");

        #endregion

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

        #region 프로그램 시작하기 - Main()

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

            MLContext context = new MLContext();

            DataOperationsCatalog.TrainTestData trainTestData = GetTrainTestData(context);

            ITransformer model = GetModel(context, trainTestData.TrainSet);

            Evaluate(context, model, trainTestData.TestSet);

            PredictSingleItem(context, model);

            Predict(context, model);

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

        #endregion

        #region 훈련 테스트 데이터 구하기 - GetTrainTestData(context)

        /// <summary>
        /// 훈련 테스트 데이터 구하기
        /// </summary>
        /// <param name="context">ML 컨텍스트</param>
        /// <returns>훈련 테스트 데이터</returns>
        public static DataOperationsCatalog.TrainTestData GetTrainTestData(MLContext context)
        {
            Console.WriteLine("BEGIN GET TRAIN TEST DATA");

            IDataView dataView = context.Data.LoadFromTextFile<SentimentData>(_labelFilePath, hasHeader : false);

            DataOperationsCatalog.TrainTestData trainTestData = context.Data.TrainTestSplit(dataView, testFraction : 0.2);

            Console.WriteLine("END GET TRAIN TEST DATA");

            return trainTestData;
        }

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

        /// <summary>
        /// 모델 구하기
        /// </summary>
        /// <param name="context">컨텍스트</param>
        /// <param name="trainingDataView">훈련 데이터 뷰</param>
        /// <returns>모델</returns>
        public static ITransformer GetModel(MLContext context, IDataView trainingDataView)
        {
            Console.WriteLine("BEGIN GET MODEL FUNCTION");

            var estimator = context.Transforms.Text.FeaturizeText
            (
                outputColumnName : "Features",
                inputColumnName  : nameof(SentimentData.SentimentText)
            )
            .Append
            (
                context.BinaryClassification.Trainers.SdcaLogisticRegression
                (
                    labelColumnName   : "Label",
                    featureColumnName : "Features"
                )
            );

            Console.WriteLine("BEGIN CREATE AND TRAIN THE MODEL");

            var model = estimator.Fit(trainingDataView);

            Console.WriteLine("END CREATE AND TRAIN THE MODEL");

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

            return model;
        }

        #endregion
        #region 평가하기 - Evaluate(context, model, testDataView)

        /// <summary>
        /// 평가하기
        /// </summary>
        /// <param name="context">ML 컨텍스트</param>
        /// <param name="model">모델</param>
        /// <param name="testDataView">테스트 데이터 뷰</param>
        public static void Evaluate(MLContext context, ITransformer model, IDataView testDataView)
        {
            Console.WriteLine("BEGIN EVALUATE FUNCTION");

            IDataView predictions = model.Transform(testDataView);

            CalibratedBinaryClassificationMetrics metrics = context.BinaryClassification.Evaluate(predictions, "Label");

            Console.WriteLine("MODEL QUALITY METRICS EVALUATION");
            Console.WriteLine("--------------------------------------------------");
            Console.WriteLine($"ACCURACY : {metrics.Accuracy:P2}"         );
            Console.WriteLine($"AUC      : {metrics.AreaUnderRocCurve:P2}");
            Console.WriteLine($"F1SCORE  : {metrics.F1Score:P2}"          );
            Console.WriteLine("--------------------------------------------------");

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

        #endregion
        #region 단일 항목 예측하기 - PredictSingleItem(context, model)

        /// <summary>
        /// 단일 항목 예측하기
        /// </summary>
        /// <param name="context">컨텍스트</param>
        /// <param name="model">모델</param>
        private static void PredictSingleItem(MLContext context, ITransformer model)
        {
            Console.WriteLine("BEGIN PREDICT SINGLE ITEM FUNCTION");

            PredictionEngine<SentimentData, SentimentPrediction> predictionEngine = context.Model.CreatePredictionEngine<SentimentData, SentimentPrediction>(model);

            SentimentData sentimentData = new SentimentData { SentimentText = "This was a very bad steak" };

            SentimentPrediction sentimentPrediction = predictionEngine.Predict(sentimentData);

            Console.WriteLine("--------------------------------------------------");
            Console.WriteLine($"SENTIMENT   : {sentimentPrediction.SentimentText}"                                            );
            Console.WriteLine($"PREDICTION  : {(Convert.ToBoolean(sentimentPrediction.Prediction) ? "Positive" : "Negative")}");
            Console.WriteLine($"PROBABILITY : {sentimentPrediction.Probability}"                                              );
            Console.WriteLine("--------------------------------------------------");

            Console.WriteLine("END PREDICT SINGLE ITEM FUNCTION");
        }

        #endregion
        #region 예측하기 - Predict(context, model)

        /// <summary>
        /// 예측하기
        /// </summary>
        /// <param name="context">ML 컨텍스트</param>
        /// <param name="model">모델</param>
        public static void Predict(MLContext context, ITransformer model)
        {
            Console.WriteLine("BEGIN PREDICT FUNCTION");

            IEnumerable<SentimentData> sentimentDataEnumerable = new[]
            {
                new SentimentData { SentimentText = "This was a horrible meal" },
                new SentimentData { SentimentText = "I love this spaghetti."   }
            };

            IDataView sentimentDataView = model.Transform(context.Data.LoadFromEnumerable(sentimentDataEnumerable));

            IEnumerable<SentimentPrediction> sentimentPredictionEnumerable = context.Data.CreateEnumerable<SentimentPrediction>(sentimentDataView, reuseRowObject : false);

            Console.WriteLine("--------------------------------------------------");

            foreach(SentimentPrediction sentimentPrediction  in sentimentPredictionEnumerable)
            {
                Console.WriteLine($"SENTIMENT   : {sentimentPrediction.SentimentText}");
                Console.WriteLine($"PREDICTION  : {(Convert.ToBoolean(sentimentPrediction.Prediction) ? "Positive" : "Negative")}");
                Console.WriteLine($"PROBABILITY : {sentimentPrediction.Probability}");
                Console.WriteLine("--------------------------------------------------");
            }

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

        #endregion
    }
}