Prediction - script API

Functions for predicting data available in StackState

Function: predictMetrics

Predict metrics for any metric query coming from any data source.


  • predictorName - the prediction preset to use:

    • fft - Fast Fourier Transformation. Used for seasonal metrics with a repeatable pattern that has a cycle.

    • linear - used for metrics that have a linear trend.

    • hmn - Harmonic Mean Normal. Used in difficult cases when metrics are not clearly seasonal or linear.

  • horizon - how much future to predict. The horizon is specified in duration format.

  • query - the metrics to use for the prediction. A telemetry query that returns metrics, created using the Telemetry.query() function followed by .compileQuery().

Builder methods:

  • predictionPoints(points: Int) - the number of points to the horizon.

  • includeHistory(start?: Instant, end?: Instant) - call this builder method to include the result of the query in the return value. Optionally a start and end can be added to limit the included history using the instant format. When not specifying the start and end the whole history will be included.

Return type:


A PredictionResponse, which contains the following fields:


  • PredictionResponse.request - the request made to the prediction API of type PredictionRequest.

  • PredictionResponse.history - optional, the history used for prediction of type MetricTelemetry. Empty if .includeHistory() was not used.

  • PredictionResponse.prediction - the predicted metrics.

The PredictionRequest type has the following fields:

  • PredictionRequest.query - the query provided to predictMetrics.

  • PredictionRequest.predictor - the name and configuration of the predictor.

  • PredictionRequest.horizon - the prediction horizon.

  • PredictionRequest.predictionPointCount - the number of predicted points.

  • PredictionRequest.historyResponse - options of the history response.

The MetricTelemetry has the following fields:

  • - the two dimensional array with values and time stamps.


Predict 8 points of disk utilization for myHost for the next four hours based on the last four weeks of data:

Prediction.predictMetrics("linear", "4h",
Telemetry.query("MyDatasource", 'host="myHost" and name="diskutil"')