Microsoft MVP성태의 닷넷 이야기
글쓴 사람
정성태 (techsharer at outlook.com)
홈페이지
첨부 파일
(연관된 글이 1개 있습니다.)

ML.NET Model Builder - 회귀(Regression), 다중 분류(Multi-class classification) 예제

지난번에 설명한,

Visual Studio - ML.NET Model Builder 소개
; https://www.sysnet.pe.kr/2/0/11894

예제는 간단한 2진 분류였는데요, 지난 튜토리얼의 마지막 단계까지 가면,

ML.NET Tutorial - Get started in 10 minutes
; https://dotnet.microsoft.com/learn/machinelearning-ai/ml-dotnet-get-started-tutorial/next

이제 또 다른 시나리오를 실습해 보라면서 "Price prediction dataset" 예제 파일을 링크하고 있습니다.

Price prediction dataset
; https://raw.githubusercontent.com/dotnet/machinelearning-samples/master/samples/csharp/getting-started/Regression_TaxiFarePrediction/TaxiFarePrediction/Data/taxi-fare-train.csv

혹시 당황하셨다면 ^^ 완벽한 예제 코드로 정리된 다음의 문서를 보면 됩니다.

dotnet/machinelearning-samples
; https://github.com/dotnet/machinelearning-samples/tree/master/samples/csharp/getting-started/Regression_TaxiFarePrediction




그러니까 결국 문제는 Model Builder를 사용할 때 어떤 종류의 기계학습에 대한 시나리오가 맞는지 선택하는 것입니다.

  1. 회귀
  2. 분류 - {2진 분류, 다중 분류}

일단, 2진 분류는 결과가 Yes/No로 나온다는 점에서 기준이 매우 간단합니다. 그리고 회귀와 다중 분류의 경우는 결과가 학습 데이터의 "Label"로 제한된 것이라면 다중 분류, 그렇지 않고 연속 공간에서 나오는 것이라면 회귀라고 간단하게 정리할 수 있습니다.

따라서, 이번 주제인 "택시 요금 예측(Taxi Fare Prediction)"은 결괏값이 연속 공간이므로 회귀에 해당합니다. 그럼 Model Builder를 간단하게 사용할 수 있겠죠. ^^

1. Scenario
    Price Prediction

2. Data
    Input: File
    Select a file: taxi-fare-train.csv
    Column to Predict (Label): fare_amount

3. Train
    Time to train (seconds): 10

MLModel.zip 파일이 생성되었으면 이전 글과 동일하게 다음의 작업을 추가하고,

1) Install-Package Microsoft.ML
   Install-Package Microsoft.ML.FastTree
2) MLModel.zip 추가 -  "Copy to Output Directory" - "Copy if newer"
3) ModelInput.cs, ModelOutput.cs 추가

예측 코드를 작성하면 됩니다.

using System;
using Microsoft.ML;
using ConsoleApp1ML.Model.DataModels;

class Program
{
    static void Main(string[] args)
    {
        MLContext mlContext = new MLContext();

        ITransformer mlModel = mlContext.Model.Load("MLModel.zip", out DataViewSchema inputSchema);
        var predEngine = mlContext.Model.CreatePredictionEngine<ModelInput, ModelOutput>(mlModel);

        // Create sample data to do a single prediction with it 
        ModelInput sampleData = new ModelInput
        {
            Vendor_id = "VTS",
            Rate_code = 1.0f,
            Passenger_count = 1,
            Trip_time_in_secs = 1140,
            Trip_distance = 3.75f,
            Payment_type = "CRD",
        };

        // Try a single prediction
        ModelOutput result = predEngine.Predict(sampleData);

        Console.WriteLine($"Single Prediction --> Predicted value: {result.Score}");
    }
}

/* 출력 결과
Single Prediction --> Predicted value: 15.95807
*/




이렇게 해서 "2진 분류"와 "회귀"에 대한 예제를 살펴봤는데요. "다중 분류"도 마저 살펴보겠습니다. 다중 분류의 가장 유명한 사례가 바로 붓꽃 판정입니다.

Iris Data Set 
; https://archive.ics.uci.edu/ml/datasets/iris

즉, 결괏값이 택시 요금 예측과 같이 연속 공간이 아니라, 데이터 파일 자체에 포함된 Label(붓꽃 데이터의 경우 class) 집합으로 한정되기 때문에 "다중 분류" 시나리오가 됩니다.

그럼 Python 예제만 있는 붓꽃 분류를 ^^ C# ML.NET으로 해보겠습니다.

우선, 위의 사이트에서 다운로드한 iris.data는 CSV 형식의 파일이지만 아쉽게도 칼럼 정보가 없습니다. 대신 iris.names 파일을 보면 다음과 같이 속성 정보가 있으니,

1. sepal length in cm 
2. sepal width in cm 
3. petal length in cm 
4. petal width in cm 
5. class: 
 -- Iris Setosa 
 -- Iris Versicolour 
 -- Iris Virginica

이를 참고해 iris.data의 첫 행에 다음과 같이 칼럼 정보를 넣고 파일명을 .csv를 붙여 저장합니다.

sepal_length,sepal_width,petal_length,petal_width,class
5.1,3.5,1.4,0.2,Iris-setosa
4.9,3.0,1.4,0.2,Iris-setosa
...[생략]...

끝입니다 이제 Model Builder를 실행해 다음과 같은 설정으로 자동 코드를 생성하고,

1. Scenario
    Custom Scenario

2. Data
    Input: File
    Select a file: iris.data.csv
    Column to Predict (Label): class

3. Train
    Machine learning task: multiclass-classification
    Time to train (seconds): 10

자동 코드가 생성되었으면 역시 우리의 응용 프로그램 프로젝트에 다음과 같은 설정을 한 후,

1) Install-Package Microsoft.ML
2) MLModel.zip 추가 -  "Copy to Output Directory" - "Copy if newer"
3) ModelInput.cs, ModelOutput.cs 추가

간단하게 붓꽃 판정 코드를 만들 수 있습니다. ^^

using ConsoleApp1ML.Model.DataModels;
using Microsoft.ML;
using System;

namespace ConsoleApp2
{
    class Program
    {
        static void Main(string[] args)
        {
            MLContext mlContext = new MLContext();

            ITransformer mlModel = mlContext.Model.Load("MLModel.zip", out DataViewSchema inputSchema);
            var predEngine = mlContext.Model.CreatePredictionEngine<ModelInput, ModelOutput>(mlModel);

            ModelInput sampleData = new ModelInput
            {
                Sepal_length = 5,
                Sepal_width = 2.9f,
                Petal_length = 1,
                Petal_width = 0.2f,
            };

            ModelOutput predictionResult = predEngine.Predict(sampleData);

            Console.WriteLine($"Single Prediction --> Predicted value: {predictionResult.Prediction} | Predicted scores: [{String.Join(",", predictionResult.Score)}]");
        }
    }
}

/* 출력 결과
Single Prediction --> Predicted value: Iris-setosa | Predicted scores: [0.8280767,0.1602236,0.01169968]
*/

엄청 쉽죠? ^^

(첨부 파일은 이 글의 예제 코드를 포함합니다.)




[이 글에 대해서 여러분들과 의견을 공유하고 싶습니다. 틀리거나 미흡한 부분 또는 의문 사항이 있으시면 언제든 댓글 남겨주십시오.]

[연관 글]






[최초 등록일: ]
[최종 수정일: 5/12/2019]

Creative Commons License
이 저작물은 크리에이티브 커먼즈 코리아 저작자표시-비영리-변경금지 2.0 대한민국 라이센스에 따라 이용하실 수 있습니다.
by SeongTae Jeong, mailto:techsharer at outlook.com

비밀번호

댓글 작성자
 




1  2  [3]  4  5  6  7  8  9  10  11  12  13  14  15  ...
NoWriterDateCnt.TitleFile(s)
13558정성태2/18/20242156닷넷: 2217. C# - 최댓값이 1인 SemaphoreSlim 보다 Mutex 또는 lock(obj)를 선택하는 것이 나은 이유
13557정성태2/18/20241908Windows: 258. Task Scheduler의 Author 속성 값을 변경하는 방법
13556정성태2/17/20241953Windows: 257. Windows - Symbolic (hard/soft) Link 및 Junction 차이점
13555정성태2/15/20242107닷넷: 2216. C# - SemaphoreSlim 사용 시 주의점
13554정성태2/15/20241851VS.NET IDE: 189. Visual Studio - 닷넷 소스코드 디컴파일 찾기가 안 될 때
13553정성태2/14/20241935닷넷: 2215. windbg - thin/fat lock 없이 동작하는 Monitor.Wait + Pulse
13552정성태2/13/20241883닷넷: 2214. windbg - Monitor.Enter의 thin lock과 fat lock
13551정성태2/12/20242079닷넷: 2213. ASP.NET/Core 웹 응용 프로그램 - 2차 스레드의 예외로 인한 비정상 종료
13550정성태2/11/20242164Windows: 256. C# - Server socket이 닫히면 Accept 시켰던 자식 소켓이 닫힐까요?
13549정성태2/3/20242495개발 환경 구성: 706. C# - 컨테이너에서 실행하기 위한 (소켓) 콘솔 프로젝트 구성
13548정성태2/1/20242322개발 환경 구성: 705. "Docker Desktop for Windows" - ASP.NET Core 응용 프로그램의 소켓 주소 바인딩(IPv4/IPv6 loopback, Any)
13547정성태1/31/20242071개발 환경 구성: 704. Visual Studio - .NET 8 프로젝트부터 dockerfile에 추가된 "USER app" 설정
13546정성태1/30/20241924Windows: 255. (디버거의 영향 등으로) 대상 프로세스가 멈추면 Socket KeepAlive로 연결이 끊길까요?
13545정성태1/30/20241845닷넷: 2212. ASP.NET Core - 우선순위에 따른 HTTP/HTTPS 호스트:포트 바인딩 방법
13544정성태1/30/20241856오류 유형: 894. Microsoft.Data.SqlClient - Could not load file or assembly 'System.Security.Permissions, ...'
13543정성태1/30/20241854Windows: 254. Windows - 기본 사용 중인 5357 포트 비활성화는 방법
13542정성태1/30/20241886오류 유형: 893. Visual Studio - Web Application을 실행하지 못하는 IISExpress - 두 번째 이야기
13541정성태1/29/20241932VS.NET IDE: 188. launchSettings.json의 useSSL 옵션
13540정성태1/29/20242061Linux: 69. 리눅스 - "Docker Desktop for Windows" Container 환경에서 IPv6 Loopback Address 바인딩 오류
13539정성태1/26/20242340개발 환경 구성: 703. Visual Studio - launchSettings.json을 이용한 HTTP/HTTPS 포트 바인딩
13538정성태1/25/20242402닷넷: 2211. C# - NonGC(FOH) 영역에 .NET 개체를 생성파일 다운로드1
13537정성태1/24/20242461닷넷: 2210. C# - Native 메모리에 .NET 개체를 생성파일 다운로드1
13536정성태1/23/20242565닷넷: 2209. .NET 8 - NonGC Heap / FOH (Frozen Object Heap) [1]
13535정성태1/22/20242406닷넷: 2208. C# - GCHandle 구조체의 메모리 분석
13534정성태1/21/20242241닷넷: 2207. C# - SQL Server DB를 bacpac으로 Export/Import파일 다운로드1
13533정성태1/18/20242453닷넷: 2206. C# - TCP KeepAlive의 서버 측 구현파일 다운로드1
1  2  [3]  4  5  6  7  8  9  10  11  12  13  14  15  ...