많은 분들의 요청에 따라, 이제 저희는 기계 학습 코드와 패키지의 예제를 제공하기 위해 열심히 작업하고 있습니다.

TensorFlow, PyTorch, Keras, Numpy, Pandas 및 cTrader 내에서 사용할 수 있는 다양한 .NET 패키지들.

저희의 목표는 cTrader 내에서 기계 학습을 모두가 더 쉽게 사용할 수 있도록 하는 것입니다.

행운을 빕니다!

*** 이 코드는 어떠한 거래도 하지 않습니다 (단지 데이터를 출력할 뿐입니다). 이는 저희 기계 학습 패키지를 사용하여 자신만의 AI 모델을 시작하는 방법에 대한 샘플 코드일 뿐입니다.

.......................................................

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using cAlgo.API;
using cAlgo.API.Collections;
using cAlgo.API.Indicators;
using cAlgo.API.Internals;

using NumSharp;
using np = NumSharp.np;
using Shape = NumSharp.Shape;

using PandasNet;
using static PandasNet.PandasApi;

namespace cAlgo.Robots
{
    [Robot(TimeZone = TimeZones.UTC, AccessRights = AccessRights.None)]
    public class E7BBKGNumSharpSample : Robot
    {
        [Parameter("Version 1.01", DefaultValue = "Version 1.01")]
        public string Version { get; set; }

        [Parameter("Source")]
        public DataSeries Source { get; set; }

        [Parameter("Bars Required", DefaultValue = 50, MinValue = 1, MaxValue = 10000, Step = 1)]
        public int BarsRequired { get; set; }

        [Parameter("Method Name", DefaultValue = MethodName.DataSplitPrints)]
        public MethodName Mode { get; set; }
        public enum MethodName
        {
            DataSplitPrints,
            PandasPrints,
            NDArrayPrints
        }
        
        protected override void OnStart()
        {
            // 지표 초기화
        }

        protected override void OnBar()
        {
            try
            {
                if (Mode == MethodName.DataSplitPrints)
                {
                    DataSplitPrints();
                }
                else if (Mode == MethodName.PandasPrints)
                {
                    PandasPrints();
                }
                else if (Mode == MethodName.NDArrayPrints)
                {
                    NDArrayPrints();
                }
            }
            catch (Exception ex)
            {
                Print($"오류: {ex.Message}");
                if (ex.InnerException != null)
                {
                    Print($"내부 예외: {ex.InnerException.Message}");
                    throw;
                }
            }
        }

        private float[,] GetDataSet()
        {
            int startBar = Bars.ClosePrices.Count - BarsRequired;
            float[,] inputSignals = new float[BarsRequired, 5];

            for (int i = 0; i < BarsRequired; i++)
            {
                int barIndex = startBar + i;
                inputSignals[i, 0] = (float)Bars.OpenPrices[barIndex];
                inputSignals[i, 1] = (float)Bars.HighPrices[barIndex];
                inputSignals[i, 2] = (float)Bars.LowPrices[barIndex];
                inputSignals[i, 3] = (float)Bars.ClosePrices[barIndex];
                inputSignals[i, 4] = (float)Bars.TickVolumes[barIndex];
            }
            return inputSignals;
        }
        
        private float[,] GetTargetDataSet()
        {
            int startBar = Bars.ClosePrices.Count - BarsRequired;
            float[,] inputSignals = new float[BarsRequired, 5];

            for (int i = 0; i < BarsRequired; i++)
            {
                int barIndex = startBar + i;
                inputSignals[i, 0] = (float)Bars.OpenPrices[barIndex];
                inputSignals[i, 1] = (float)Bars.HighPrices[barIndex];
                inputSignals[i, 2] = (float)Bars.LowPrices[barIndex];
                inputSignals[i, 3] = (float)Bars.ClosePrices[barIndex];
                inputSignals[i, 4] = (float)Bars.TickVolumes[barIndex];
            }
            return inputSignals;
        }
        
        /// NumSharp 데이터 분할 출력
        public void DataSplitPrints()
        {
            // 모델이 예상하는 입력 형태에 맞게 입력 데이터를 재구성
            //var inputShape = new Shape(-1, BarsRequired, 5);
            NDArray inputData = np.array<float>(GetDataSet());
            Print("입력 NDarray: " + string.Join(", ", inputData));
            
            // 모델이 예상하는 대상 형태에 맞게 대상 데이터를 재구성
            //var targetShape = new Shape(-1, 5);
            NDArray targetData = np.array<float>(GetTargetDataSet());
            Print("대상 NDarray: " + string.Join(", ", targetData));
            
            // 데이터를 훈련 세트와 테스트 세트로 분할
            int testSize = (int)(0.2 * inputData.shape[0]); // 테스트용 20%
            var (x_train, x_test) = (inputData[$":{inputData.shape[0] - testSize}"], inputData[$"{inputData.shape[0] - testSize}:"]);
            var (y_train, y_test) = (targetData[$":{targetData.shape[0] - testSize}"], targetData[$"{targetData.shape[0] - testSize}:"]);
            
            Print("X_train 데이터: " + string.Join(", ", x_train));
            Print("X_test 데이터: " + string.Join(", ", x_test));
            Print("Y_train 데이터: " + string.Join(", ", y_train));
            Print("Y_test 데이터: " + string.Join(", ", y_test));
        }
        
        /// PandasNet 출력
        public void PandasPrints()
        {
            // float[,]를 List<Series>로 변환
            var inputData = GetDataSet();
            var targetData = GetTargetDataSet();
            
            var inputSeriesList = new List<Series>();
            var targetSeriesList = new List<Series>();
            
            for (int col = 0; col < inputData.GetLength(1); col++)
            {
                List<float> columnData = new List<float>();
                for (int row = 0; row < inputData.GetLength(0); row++)
                {
                    columnData.Add(inputData[row, col]);
                }
                inputSeriesList.Add(new Series(columnData.ToArray()));
            }
            
            for (int col = 0; col < targetData.GetLength(1); col++)
            {
                List<float> columnData = new List<float>();
                for (int row = 0; row < targetData.GetLength(0); row++)
                {
                    columnData.Add(targetData[row, col]);
                }
                targetSeriesList.Add(new Series(columnData.ToArray()));
            }
            
            // 데이터프레임 생성
            DataFrame inputDataFrame = new DataFrame(inputSeriesList);
            DataFrame targetDataFrame = new DataFrame(targetSeriesList);
            
            Print("입력 데이터프레임: " + inputDataFrame);
            Print("대상 데이터프레임: " + targetDataFrame);
            
            //Print("입력 데이터프레임: " + string.Join(", ", inputDataFrame));
            //Print("대상 데이터프레임: " + string.Join(", ", targetDataFrame));
        }
        
        /// 간단한 NumSharp NDArray 출력
        public void NDArrayPrints()
        {
            if (Bars.ClosePrices.Count < BarsRequired)
                return;

            try
            {
                // 입력 데이터 float[,] 호출
                float[,] inputData = GetDataSet();

                // NDArray로 변환하고 (BarsRequired, 5) 형태로 재구성
                NDArray inputNDArray = np.array(inputData);   // NumSharp
                Print("입력 NumSharp NDarray 데이터 : " + string.Join(", ", inputNDArray));
                Print("입력 NumSharp NDarray 형태: " + string.Join(", ", inputNDArray.shape));
                
                int expectedLength = BarsRequired * 5;
                Print($"예상 NumSharp NDarray 길이: {expectedLength}");
                Print($"입력 NumSharp NDarray 크기: {inputNDArray.size}");

                if (inputNDArray.size != expectedLength)
                {
                    Print($"길이 불일치: 예상 길이 {expectedLength}, 실제 크기 {inputNDArray.size}");
                    return;
                }
            }
            catch (Exception ex)
            {
                Print("예외: " + ex.Message);
                Print("스택 추적: " + ex.StackTrace);

                Exception innerException = ex.InnerException;
                while (innerException != null)
                {
                    Print("내부 예외: " + innerException.Message);
                    Print("내부 예외 스택 추적: " + innerException.StackTrace);
                    innerException = innerException.InnerException;
                }
            }
        }
    }
}