Русский RU
Открыть на GitHub

Кластер K-Means на основе Боллинджера

Стратегия Bollinger K-Means Cluster построена на использовании полос Боллинджера совместно с кластеризацией K-Means.

Сигналы формируются, когда Боллинджер подтверждает изменения тренда на внутридневных данных (5м). Такой метод подходит активным трейдерам.

Стопы рассчитываются на основе кратных ATR и параметров BollingerLength, BollingerDeviation. Настройте эти значения для баланса риска и прибыли.

Подробности

  • Критерии входа: см. реализацию условий индикаторов.
  • Длинные/короткие: обе стороны.
  • Критерии выхода: противоположный сигнал или логика стопов.
  • Стопы: да, расчёт на основе индикаторов.
  • Значения по умолчанию:
    • BollingerLength = 20
    • BollingerDeviation = 2.0m
    • CandleType = TimeSpan.FromMinutes(5).TimeFrame()
    • KMeansHistoryLength = 50
  • Фильтры:
    • Категория: Следование за трендом
    • Направление: Оба
    • Индикаторы: Bollinger
    • Стопы: Да
    • Сложность: Средняя
    • Таймфрейм: Внутридневной (5м)
    • Сезонность: Нет
    • Нейронные сети: Нет
    • Дивергенция: Нет
    • Уровень риска: Средний
using System;
using System.Collections.Generic;

using Ecng.Common;
using Ecng.Serialization;

using StockSharp.Algo.Indicators;
using StockSharp.Algo.Strategies;
using StockSharp.BusinessEntities;
using StockSharp.Messages;

namespace StockSharp.Samples.Strategies;

/// <summary>
/// Bollinger Bands with K-Means clustering strategy.
/// Uses Bollinger Bands indicator along with a simple K-Means clustering algorithm
/// to identify overbought/oversold conditions.
/// </summary>
public class BollingerKMeansStrategy : Strategy
{
	private readonly StrategyParam<int> _bollingerLength;
	private readonly StrategyParam<decimal> _bollingerDeviation;
	private readonly StrategyParam<DataType> _candleType;
	private readonly StrategyParam<int> _kMeansHistoryLength;

	private BollingerBands _bollinger;
	private decimal _atrValue;

	// Cluster state tracking
	private enum ClusterStates
	{
		Oversold,
		Neutral,
		Overbought
	}

	private ClusterStates _currentClusterState = ClusterStates.Neutral;
	private readonly List<decimal> _rsiValues = [];
	private readonly List<decimal> _priceValues = [];
	private readonly List<decimal> _volumeValues = [];

	private RelativeStrengthIndex _rsi;
	private AverageTrueRange _atr;

	/// <summary>
	/// Bollinger Bands period.
	/// </summary>
	public int BollingerLength
	{
		get => _bollingerLength.Value;
		set => _bollingerLength.Value = value;
	}

	/// <summary>
	/// Bollinger Bands standard deviation multiplier.
	/// </summary>
	public decimal BollingerDeviation
	{
		get => _bollingerDeviation.Value;
		set => _bollingerDeviation.Value = value;
	}

	/// <summary>
	/// Candle type to use for the strategy.
	/// </summary>
	public DataType CandleType
	{
		get => _candleType.Value;
		set => _candleType.Value = value;
	}

	/// <summary>
	/// Length of history for K-Means clustering.
	/// </summary>
	public int KMeansHistoryLength
	{
		get => _kMeansHistoryLength.Value;
		set => _kMeansHistoryLength.Value = value;
	}

	/// <summary>
	/// Initializes a new instance of the <see cref="BollingerKMeansStrategy"/>.
	/// </summary>
	public BollingerKMeansStrategy()
	{
		_bollingerLength = Param(nameof(BollingerLength), 20)
		.SetDisplay("Bollinger Length", "Length of the Bollinger Bands indicator", "Indicators")
		
		.SetOptimize(10, 50, 5);

		_bollingerDeviation = Param(nameof(BollingerDeviation), 2.0m)
		.SetDisplay("Bollinger Deviation", "Standard deviation multiplier for Bollinger Bands", "Indicators")
		
		.SetOptimize(1.0m, 3.0m, 0.5m);

		_candleType = Param(nameof(CandleType), TimeSpan.FromMinutes(5).TimeFrame())
		.SetDisplay("Candle Type", "Type of candles to use", "General");

		_kMeansHistoryLength = Param(nameof(KMeansHistoryLength), 50)
		.SetDisplay("K-Means History Length", "Length of history for K-Means clustering", "Clustering")
		
		.SetOptimize(30, 100, 10);
	}

	/// <inheritdoc />
	public override IEnumerable<(Security sec, DataType dt)> GetWorkingSecurities()
	{
		return [(Security, CandleType)];
	}

	/// <inheritdoc />
	protected override void OnReseted()
	{
		base.OnReseted();

		_atrValue = default;
_currentClusterState = ClusterStates.Neutral;
		_rsiValues.Clear();
		_priceValues.Clear();
		_volumeValues.Clear();

		_bollinger?.Reset();
		_rsi?.Reset();
		_atr?.Reset();
	}

	/// <inheritdoc />
	protected override void OnStarted2(DateTime time)
	{
		base.OnStarted2(time);

		// Create indicators
		_bollinger = new BollingerBands
		{
			Length = BollingerLength,
			Width = BollingerDeviation
		};

		_rsi = new RelativeStrengthIndex
		{
			Length = 14
		};

		_atr = new AverageTrueRange
		{
			Length = 14
		};

		// Create and initialize subscription
		var subscription = SubscribeCandles(CandleType);

		subscription
		.BindEx(
		_bollinger, 
		_rsi,
		_atr,
		ProcessCandle)
		.Start();

		// Setup chart visualization if available
		var area = CreateChartArea();
		if (area != null)
		{
			DrawCandles(area, subscription);
			DrawIndicator(area, _bollinger);
			DrawOwnTrades(area);
		}

		// Setup position protection
		StartProtection(
		new Unit(2, UnitTypes.Percent), 
		new Unit(2, UnitTypes.Percent)
		);
	}

	private void ProcessCandle(ICandleMessage candle, IIndicatorValue bollingerValue, IIndicatorValue rsiValue, IIndicatorValue atrValue)
	{
		// Skip unfinished candles
		if (candle.State != CandleStates.Finished)
		return;

		// Check if strategy is ready to trade
		if (!IsFormedAndOnlineAndAllowTrading())
		return;

		if (bollingerValue is not IBollingerBandsValue bollingerTyped ||
			bollingerTyped.UpBand is not decimal bollingerUpper ||
			bollingerTyped.MovingAverage is not decimal bollingerMiddle ||
			bollingerTyped.LowBand is not decimal bollingerLower)
			return;

		var rsi = rsiValue.ToDecimal();
		_atrValue = atrValue.ToDecimal();

		// Update data for clustering
		UpdateClusterData(candle, rsi);

		// Calculate K-Means clusters and determine market state
		CalculateClusters();

		// Use ATR as a minimal band breach filter to avoid noise around the envelopes.
		var bandBuffer = Math.Max(_atrValue * 0.1m, Security?.PriceStep ?? 0m);

		// Trading logic
		if (candle.ClosePrice < bollingerLower - bandBuffer && _currentClusterState == ClusterStates.Oversold && Position <= 0)
		{
			// Buy signal - price below lower band and in oversold cluster
			BuyMarket(Volume);
			LogInfo($"Buy Signal: Price below lower band ({bollingerLower:F2}) in oversold cluster");
		}
		else if (candle.ClosePrice > bollingerUpper + bandBuffer && _currentClusterState == ClusterStates.Overbought && Position >= 0)
		{
			// Sell signal - price above upper band and in overbought cluster
			SellMarket(Volume + Math.Abs(Position));
			LogInfo($"Sell Signal: Price above upper band ({bollingerUpper:F2}) in overbought cluster");
		}
		else if (Position > 0 && candle.ClosePrice > bollingerMiddle)
		{
			// Exit long position when price returns to middle band
			SellMarket(Position);
			LogInfo($"Exit Long: Price returned to middle band ({bollingerMiddle:F2})");
		}
		else if (Position < 0 && candle.ClosePrice < bollingerMiddle)
		{
			// Exit short position when price returns to middle band
			BuyMarket(Math.Abs(Position));
			LogInfo($"Exit Short: Price returned to middle band ({bollingerMiddle:F2})");
		}
	}

	private void UpdateClusterData(ICandleMessage candle, decimal rsi)
	{
		// Add current values to the data series
		_priceValues.Add(candle.ClosePrice);
		_rsiValues.Add(rsi);
		_volumeValues.Add(candle.TotalVolume);

		// Maintain the desired history length
		while (_priceValues.Count > KMeansHistoryLength)
		{
			_priceValues.RemoveAt(0);
			_rsiValues.RemoveAt(0);
			_volumeValues.RemoveAt(0);
		}
	}

	private void CalculateClusters()
	{
		// Only perform clustering when we have enough data
		if (_priceValues.Count < KMeansHistoryLength)
		return;

		// Normalize the data (simple min-max normalization)
		var priceValues = _priceValues.ToArray();
		var rsiValues = _rsiValues.ToArray();
		var normalizedRsi = rsiValues[^1] / 100m;  // RSI is already 0-100

		// Find min/max for price normalization
		decimal? minPrice = null;
		decimal? maxPrice = null;

		foreach (var price in priceValues)
		{
			if (minPrice == null || price < minPrice.Value)
			minPrice = price;
			if (maxPrice == null || price > maxPrice.Value)
			maxPrice = price;
		}

		// Normalize the last price
		decimal normalizedPrice = 0.5m;
		if (minPrice.HasValue && maxPrice.HasValue && maxPrice.Value != minPrice.Value)
		{
			var priceRange = maxPrice.Value - minPrice.Value;
			normalizedPrice = (priceValues[^1] - minPrice.Value) / priceRange;
		}

		// Simple rules-based clustering (simplified K-means approximation)
		// Oversold: Low RSI (< 30) and price near bottom of range
		// Overbought: High RSI (> 70) and price near top of range
		// Neutral: Everything else

		if (normalizedRsi < 0.3m && normalizedPrice < 0.3m)
		{
			_currentClusterState = ClusterStates.Oversold;
		}
		else if (normalizedRsi > 0.7m && normalizedPrice > 0.7m)
		{
			_currentClusterState = ClusterStates.Overbought;
		}
		else
		{
			_currentClusterState = ClusterStates.Neutral;
		}

		LogInfo($"Cluster State: {_currentClusterState}, Normalized RSI: {normalizedRsi:F2}, Normalized Price: {normalizedPrice:F2}");
	}
}