using System.Runtime.CompilerServices; namespace QuanTAlib; /// /// MDA: Mean Directional Accuracy /// A metric that measures how well a forecast predicts the direction of change /// rather than the magnitude. MDA focuses on whether the predicted movement /// (up or down) matches the actual movement. /// /// /// The MDA calculation process: /// 1. For each consecutive pair of points: /// - Calculate direction of actual change /// - Calculate direction of predicted change /// - Compare directions (match = 1, mismatch = 0) /// 2. Average the directional matches /// /// Key characteristics: /// - Scale-independent (only considers direction) /// - Range is 0 to 1 (easy interpretation) /// - Useful for trend prediction evaluation /// - Ignores magnitude of changes /// - Equal weight to all directional changes /// /// Formula: /// MDA = (1/(n-1)) * Σ(sign(actual[t] - actual[t-1]) == sign(pred[t] - pred[t-1])) /// /// Sources: /// https://www.sciencedirect.com/science/article/abs/pii/S0169207016000121 /// "Evaluating Forecasting Performance" - International Journal of Forecasting /// [SkipLocalsInit] public sealed class Mda : AbstractBase { private readonly CircularBuffer _actualBuffer; private readonly CircularBuffer _predictedBuffer; /// The number of points over which to calculate the MDA. /// Thrown when period is less than 1. [MethodImpl(MethodImplOptions.AggressiveInlining)] public Mda(int period) { if (period < 1) { throw new ArgumentOutOfRangeException(nameof(period), "Period must be greater than or equal to 1."); } WarmupPeriod = period; _actualBuffer = new CircularBuffer(period); _predictedBuffer = new CircularBuffer(period); Name = $"Mda(period={period})"; Init(); } /// The data source object that publishes updates. /// The number of points over which to calculate the MDA. [MethodImpl(MethodImplOptions.AggressiveInlining)] public Mda(object source, int period) : this(period) { var pubEvent = source.GetType().GetEvent("Pub"); pubEvent?.AddEventHandler(source, new ValueSignal(Sub)); } [MethodImpl(MethodImplOptions.AggressiveInlining)] public override void Init() { base.Init(); _actualBuffer.Clear(); _predictedBuffer.Clear(); } [MethodImpl(MethodImplOptions.AggressiveInlining)] protected override void ManageState(bool isNew) { if (isNew) { _lastValidValue = Input.Value; _index++; } } [MethodImpl(MethodImplOptions.AggressiveInlining | MethodImplOptions.AggressiveOptimization)] private static int CompareDirections(double current, double previous) { return Math.Sign(current - previous); } [MethodImpl(MethodImplOptions.AggressiveInlining | MethodImplOptions.AggressiveOptimization)] protected override double Calculation() { ManageState(Input.IsNew); double actual = Input.Value; _actualBuffer.Add(actual, Input.IsNew); // If no predicted value provided, use mean of actual values double predicted = double.IsNaN(Input2.Value) ? _actualBuffer.Average() : Input2.Value; _predictedBuffer.Add(predicted, Input.IsNew); double mda = 0; if (_actualBuffer.Count > 0) { ReadOnlySpan actualValues = _actualBuffer.GetSpan(); ReadOnlySpan predictedValues = _predictedBuffer.GetSpan(); double sumDirectionalAccuracy = 0; for (int i = 1; i < actualValues.Length; i++) { int actualDirection = CompareDirections(actualValues[i], actualValues[i - 1]); int predictedDirection = CompareDirections(predictedValues[i], predictedValues[i - 1]); sumDirectionalAccuracy += (actualDirection == predictedDirection) ? 1 : 0; } mda = sumDirectionalAccuracy / (actualValues.Length - 1); } IsHot = _index >= WarmupPeriod; return mda; } }