説明
多くの方からのご要望により、現在、当社の機械学習コードやパッケージのいくつかの例を提供するために懸命に取り組んでいます。
TensorFlow、PyTorch、Keras、Numpy、Pandas、そして多くの.NETパッケージをcTrader内で利用開始できます。
私たちの使命は、cTrader内での機械学習を誰にとってもより簡単にすることです。
ハッピー・ハンティング！
*** このコードは何も取引しません（データを出力するだけです）。これは、当社の機械学習パッケージを使用して独自のAIモデルを作成し始める方法のサンプルコードに過ぎません。
.......................................................
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()
{
// インジケーターを初期化します
}
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($"エラー: {ex.Message}");
if (ex.InnerException != null)
{
Print($"内部例外: {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 データ分割プリント
public void DataSplitPrints()
{
// モデルの期待する入力形状に合わせて入力データをリシェイプ
//var inputShape = new Shape(-1, BarsRequired, 5);
NDArray inputData = np.array<float>(GetDataSet());
Print("入力NDarray: " + string.Join(", ", inputData));
// モデルが期待するターゲット形状に合わせてターゲットデータをリシェイプ
//var targetShape = new Shape(-1, 5);
NDArray targetData = np.array<float>(GetTargetDataSet());
Print("ターゲットNDarray: " + string.Join(", ", targetData));
// データをトレーニングセットとテストセットに分割
int testSize = (int)(0.2 * inputData.shape[0]); // テスト用に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 データ: " + string.Join(", ", x_train));
Print("X_test データ: " + string.Join(", ", x_test));
Print("Y_train データ: " + string.Join(", ", y_train));
Print("Y_test データ: " + string.Join(", ", y_test));
}
/// PandasNet プリント
public void PandasPrints()
{
// float[,] を List<Series> に変換
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()));
}
// DataFrameを作成
DataFrame inputDataFrame = new DataFrame(inputSeriesList);
DataFrame targetDataFrame = new DataFrame(targetSeriesList);
Print("入力DataFrame: " + inputDataFrame);
Print("ターゲットDataFrame: " + targetDataFrame);
//Print("Input DataFrame: " + string.Join(", ", inputDataFrame));
//Print("Target DataFrame: " + string.Join(", ", targetDataFrame));
}
/// シンプルなNumSharp NDArraysプリント
public void NDArrayPrints()
{
if (Bars.ClosePrices.Count < BarsRequired)
return;
try
{
// 入力データ float[,] を呼び出す
float[,] inputData = GetDataSet();
// NDArrayに変換し (BarsRequired, 5) にリシェイプ
NDArray inputNDArray = np.array(inputData); // NumSharp
Print("入力NumSharp NDarrayデータ : " + string.Join(", ", inputNDArray));
Print("入力NumSharp NDarray形状: " + string.Join(", ", inputNDArray.shape));
int expectedLength = BarsRequired * 5;
Print($"期待されるNumSharp NDarrayの長さ: {expectedLength}");
Print($"入力NumSharp NDarrayのサイズ: {inputNDArray.size}");
if (inputNDArray.size != expectedLength)
{
Print($"長さ不一致: 期待長さ {expectedLength}、しかしサイズは {inputNDArray.size} でした");
return;
}
}
catch (Exception ex)
{
Print("例外: " + ex.Message);
Print("スタックトレース: " + ex.StackTrace);
Exception innerException = ex.InnerException;
while (innerException != null)
{
Print("内部例外: " + innerException.Message);
Print("内部例外スタックトレース: " + innerException.StackTrace);
innerException = innerException.InnerException;
}
}
}
}
}
概要
Key functionalities include:
- Data extraction from market bars (open, high, low, close prices, and tick volumes) over a configurable number of bars.
- Conversion of this data into formats compatible with machine learning workflows, including NumSharp NDArrays and Pandas DataFrames.
- Methods to split data into training and testing sets, and to print these datasets for inspection.
- Three operational modes selectable via parameters: DataSplitPrints, PandasPrints, and NDArrayPrints, each demonstrating different data handling approaches.
This sample code aims to facilitate machine learning development inside cTrader by providing foundational examples of data preparation and manipulation using popular ML libraries in a .NET context.
![「[Fx4U] AUDCHF - Price Action」ロゴ](https://cdn.ctrader.com/image/webp/0a22e9aa-c83b-42cd-b0b8-c1a2da593207_26148)
