Implementation:Dotnet Machinelearning SsaForecastingEstimator Fit

Knowledge Sources	ML.NET ML.NET API Reference
Domains	Machine Learning, Time Series, .NET
Last Updated	2026-02-09 00:00 GMT

Overview

The Fit method on SsaForecastingEstimator trains the SSA model on a provided IDataView and returns a stateful SsaForecastingTransformer containing the trained recurrence model.

Description

SsaForecastingEstimator.Fit consumes the training data and produces a SsaForecastingTransformer. Internally, it extracts the time series values from the input column, constructs the trajectory matrix, performs SVD, selects the signal rank, derives the recurrence coefficients, and packages the result into a transformer that implements IStatefulTransformer.

The returned SsaForecastingTransformer wraps an AdaptiveSingularSpectrumSequenceModelerInternal object, which contains:

The trained SVD decomposition
The recurrence coefficients for forecasting
The internal buffer holding recent observations
The confidence interval estimation state

SsaForecastingTransformer extends SsaForecastingBaseWrapper which extends TimeSeriesTransformBase<float, VBuffer<float>>, inheriting the stateful transformer infrastructure. This class hierarchy ensures proper state management for sequential data processing.

The transformer is not used directly via .Transform() for forecasting. Instead, it is wrapped in a TimeSeriesPredictionEngine which provides the stateful prediction interface with support for adaptive updates and variable-horizon forecasting.

Usage

Call .Fit(dataView) on the configured estimator. The IDataView must contain a float column matching the inputColumnName specified during estimator configuration. After fitting, use transformer.CreateTimeSeriesEngine<TSrc, TDst>(mlContext) to create the prediction engine.

Code Reference

Source Location

Repository: ML.NET
File: src/Microsoft.ML.TimeSeries/SSaForecasting.cs:L318-321 (Fit method)
File: src/Microsoft.ML.TimeSeries/SSaForecasting.cs:L29-204 (SsaForecastingTransformer)
File: src/Microsoft.ML.TimeSeries/SsaForecastingBase.cs:L19-314 (base class)

Signature

// SsaForecastingEstimator.Fit
public SsaForecastingTransformer Fit(IDataView input)

Import

using Microsoft.ML;
using Microsoft.ML.Transforms.TimeSeries;

I/O Contract

Inputs

Name	Type	Required	Description
input	IDataView	Yes	Data view containing the time series column. The column specified by inputColumnName (from the estimator configuration) must be of type `Single` (float). The first trainSize rows are used for SVD fitting.

Outputs

Name	Type	Description
(return)	SsaForecastingTransformer	A stateful transformer implementing `IStatefulTransformer`. Contains the trained AdaptiveSingularSpectrumSequenceModelerInternal with SVD results, recurrence coefficients, and internal buffer. Produces forecast values, and optionally confidence bounds, as `VBuffer<float>` output columns.

Usage Examples

Basic Example

using Microsoft.ML;
using Microsoft.ML.Transforms.TimeSeries;

var mlContext = new MLContext(seed: 42);

// Define input schema
public class TimeSeriesInput
{
    public float Value { get; set; }
}

// Load data
IDataView dataView = mlContext.Data.LoadFromTextFile<TimeSeriesInput>(
    path: "daily_values.csv",
    separatorChar: ',',
    hasHeader: true);

// Configure the SSA estimator
var estimator = mlContext.Forecasting.ForecastBySsa(
    outputColumnName: "Forecast",
    inputColumnName: "Value",
    windowSize: 7,
    seriesLength: 30,
    trainSize: 365,
    horizon: 7,
    confidenceLowerBoundColumn: "LowerBound",
    confidenceUpperBoundColumn: "UpperBound");

// Fit the model — performs SVD and builds recurrence model
SsaForecastingTransformer model = estimator.Fit(dataView);

// The model is now ready to be wrapped in a TimeSeriesPredictionEngine

Full Pipeline Example

// Fit and immediately create a prediction engine
var model = estimator.Fit(dataView);

var forecastEngine = model.CreateTimeSeriesEngine<TimeSeriesInput, ForecastOutput>(mlContext);
// forecastEngine is ready for stateful prediction

Dependencies

Microsoft.ML.TimeSeries NuGet package

Related Pages

Implements Principle

Principle:Dotnet_Machinelearning_SSA_Model_Fitting

Requires Environment

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment