QuantileForecastResult<T>

Models & Types · AiDotNet.Finance.Forecasting

Represents the result of a probabilistic/quantile forecast from a time series foundation model.

For Beginners

A quantile forecast doesn’t just predict a single value — it predicts a range of possible values with associated probabilities. For example:

  • 10th percentile: There’s a 10% chance the actual value will be below this
  • 50th percentile (median): The “middle” forecast — equally likely to be above or below
  • 90th percentile: There’s a 90% chance the actual value will be below this

This gives you prediction intervals (confidence bands) instead of just point predictions.

How It Works

Supported Models: Chronos-2, Moirai 2.0, TimesFM 2.5, Chronos-Bolt, Lag-Llama, and diffusion-based models (TimeGrad, CSDI, TSDiff) that produce sample-based forecasts.

Constructors

Constructor Summary
QuantileForecastResult(IReadOnlyList<Tensor<T>> samples, double[] quantileLevels) Creates a quantile forecast result from sample trajectories.
QuantileForecastResult(Tensor<T> pointForecast, Dictionary<double, Tensor<T>> quantileForecasts) Creates a quantile forecast result from sample trajectories.

Properties

Property Summary
IReadOnlyDictionary<double, Tensor<T>> QuantileForecasts { get; } Gets the quantile forecasts as a dictionary mapping quantile level to forecast tensor.
IReadOnlyList<double> QuantileLevels { get; } Gets the quantile levels (e.g., [0.1, 0.25, 0.5, 0.75, 0.9]).
Tensor<T> PointForecast { get; } Gets the point forecast (median or mean prediction) of shape [horizon].
int Horizon { get; } Gets the forecast horizon (number of predicted time steps).

Methods

Method Summary
ValueTuple<Tensor<T>, Tensor<T>> GetPredictionInterval(double confidenceLevel) Gets the prediction interval at the specified confidence level.