public IList <double> Execute(IList <double> source)
{
var count = source.Count;
if (count == 0)
{
return(new double[0]);
}
var res = new double[count];
var context = Context;
Contract.Assert(context != null);
var size = count / 2;
var data = source.Take(size).Select(PredictionData.MakePredictionData);
var ml = new MLContext();
var dataView = ml.Data.LoadFromEnumerable(data);
const string outputColumnName = "Prediction";
const string inputColumnName = "Value";
var args = new SsaChangePointDetector.Arguments()
{
Source = inputColumnName,
Name = outputColumnName,
Confidence = Confidence, // The confidence for spike detection in the range [0, 100]
ChangeHistoryLength = size / 4, // The length of the sliding window on p-values for computing the martingale score.
TrainingWindowSize = size / 2, // The number of points from the beginning of the sequence used for training.
SeasonalWindowSize = size / 8, // An upper bound on the largest relevant seasonality in the input time - series."
};
// Train the change point detector.
ITransformer model = new SsaChangePointEstimator(ml, args).Fit(dataView);
// Create a prediction engine from the model for feeding new data.
var engine = model.CreateTimeSeriesPredictionFunction <PredictionData, ChangePointPrediction>(ml);
for (var i = size; i < count; i++)
{
var prediction = engine.Predict(PredictionData.MakePredictionData(source[i]));
res[i] = prediction.Prediction[OutputNumber];
}
return(res);
}
// This example shows change point detection as above, but demonstrates how to train a model
// that can run predictions on streaming data, and how to persist the trained model and then re-load it.
public static void SsaChangePointDetectorPrediction()
{
// Create a new ML context, for ML.NET operations. It can be used for exception tracking and logging,
// as well as the source of randomness.
var ml = new MLContext();
// Generate sample series data with a recurring pattern
const int SeasonalitySize = 5;
const int TrainingSeasons = 3;
const int TrainingSize = SeasonalitySize * TrainingSeasons;
var data = new List <SsaChangePointData>();
for (int i = 0; i < TrainingSeasons; i++)
{
for (int j = 0; j < SeasonalitySize; j++)
{
data.Add(new SsaChangePointData(j));
}
}
// Convert data to IDataView.
var dataView = ml.CreateStreamingDataView(data);
// Setup SsaChangePointDetector arguments
var inputColumnName = nameof(SsaChangePointData.Value);
var outputColumnName = nameof(ChangePointPrediction.Prediction);
var args = new SsaChangePointDetector.Arguments()
{
Source = inputColumnName,
Name = outputColumnName,
Confidence = 95, // The confidence for spike detection in the range [0, 100]
ChangeHistoryLength = 8, // The length of the window for detecting a change in trend; shorter windows are more sensitive to spikes.
TrainingWindowSize = TrainingSize, // The number of points from the beginning of the sequence used for training.
SeasonalWindowSize = SeasonalitySize + 1 // An upper bound on the largest relevant seasonality in the input time series."
};
// Train the change point detector.
ITransformer model = new SsaChangePointEstimator(ml, args).Fit(dataView);
// Create a prediction engine from the model for feeding new data.
var engine = model.CreateTimeSeriesPredictionFunction <SsaChangePointData, ChangePointPrediction>(ml);
// Start streaming new data points with no change point to the prediction engine.
Console.WriteLine($"Output from ChangePoint predictions on new data:");
Console.WriteLine("Data\tAlert\tScore\tP-Value\tMartingale value");
ChangePointPrediction prediction = null;
for (int i = 0; i < 5; i++)
{
var value = i;
prediction = engine.Predict(new SsaChangePointData(value));
Console.WriteLine("{0}\t{1}\t{2:0.00}\t{3:0.00}\t{4:0.00}", value, prediction.Prediction[0], prediction.Prediction[1], prediction.Prediction[2], prediction.Prediction[3]);
}
// Now stream data points that reflect a change in trend.
for (int i = 0; i < 5; i++)
{
var value = (i + 1) * 100;
prediction = engine.Predict(new SsaChangePointData(value));
Console.WriteLine("{0}\t{1}\t{2:0.00}\t{3:0.00}\t{4:0.00}", value, prediction.Prediction[0], prediction.Prediction[1], prediction.Prediction[2], prediction.Prediction[3]);
}
// Now we demonstrate saving and loading the model.
// Save the model that exists within the prediction engine.
// The engine has been updating this model with every new data point.
var modelPath = "model.zip";
engine.CheckPoint(ml, modelPath);
// Load the model.
using (var file = File.OpenRead(modelPath))
model = TransformerChain.LoadFrom(ml, file);
// We must create a new prediction engine from the persisted model.
engine = model.CreateTimeSeriesPredictionFunction <SsaChangePointData, ChangePointPrediction>(ml);
// Run predictions on the loaded model.
for (int i = 0; i < 5; i++)
{
var value = (i + 1) * 100;
prediction = engine.Predict(new SsaChangePointData(value));
Console.WriteLine("{0}\t{1}\t{2:0.00}\t{3:0.00}\t{4:0.00}", value, prediction.Prediction[0], prediction.Prediction[1], prediction.Prediction[2], prediction.Prediction[3]);
}
// Output from ChangePoint predictions on new data:
// Data Alert Score P-Value Martingale value
// 0 0 - 1.01 0.50 0.00
// 1 0 - 0.24 0.22 0.00
// 2 0 - 0.31 0.30 0.00
// 3 0 0.44 0.01 0.00
// 4 0 2.16 0.00 0.24
// 100 0 86.23 0.00 2076098.24
// 200 0 171.38 0.00 809668524.21
// 300 1 256.83 0.01 22130423541.93 <-- alert is on, note that delay is expected
// 400 0 326.55 0.04 241162710263.29
// 500 0 364.82 0.08 597660527041.45 <-- saved to disk
// 100 0 - 58.58 0.15 1096021098844.34 <-- loaded from disk and running new predictions
// 200 0 - 41.24 0.20 97579154688.98
// 300 0 - 30.61 0.24 95319753.87
// 400 0 58.87 0.38 14.24
// 500 0 219.28 0.36 0.05
}
