ForecastParams

JSON representation
Period

Parameters that control the sensitivity and other options for the time series forecast.

JSON representation
{ "seasonalityHint": enum (`Period`), "horizonDuration": string, "noiseThreshold": number }

Fields

Fields
`seasonalityHint`	`enum (Period)` Optional. Specifying any known seasonality/periodicity in the time series for the slices we will analyze can improve the quality of the results. If unsure, simply leave it unspecified by not setting a value for this field. If your time series has multiple seasonal patterns, then set it to the most granular one (e.g. if it has daily and weekly patterns, set this to DAILY).
`horizonDuration`	`string (Duration format)` Optional. The length of the returned [forecasted timeseries][EvaluatedSlice.forecast]. This duration is currently capped at 100 x `granularity`. Example: If the detection point is set to "2020-12-27T00:00:00Z", the `granularity` to "3600s" and the horizonDuration to "10800s", then we will generate 3 time series points (from "2020-12-27T01:00:00Z" to "2020-12-27T04:00:00Z"), for which we will return their forecasted values. Note: The horizon time is only used for forecasting not for anormaly detection. To detect anomalies for multiple points of time, simply send multiple queries with those as `detectionTime`. A duration in seconds with up to nine fractional digits, ending with '`s`'. Example: `"3.5s"`.
`noiseThreshold`	`number` Optional. Penalize variations between the actual and forecasted values smaller than this. For more information about how this parameter affects the score, see the `anomalyScore` formula. Intuitively, anomaly scores summarize how statistically significant the change between the actual and forecasted value is compared with what we expect the change to be (see `expectedDeviation`). However, in practice, depending on the application, changes smaller than certain absolute values, while statistically significant, may not be important. This parameter allows us to penalize such low absolute value changes. Must be in the (0.0, inf) range. If unspecified, it defaults to 0.000001.

seasonalityHint

enum (Period)

Optional. Specifying any known seasonality/periodicity in the time series for the slices we will analyze can improve the quality of the results.

If unsure, simply leave it unspecified by not setting a value for this field.

If your time series has multiple seasonal patterns, then set it to the most granular one (e.g. if it has daily and weekly patterns, set this to DAILY).

horizonDuration

string (Duration format)

Optional. The length of the returned [forecasted timeseries][EvaluatedSlice.forecast].

This duration is currently capped at 100 x granularity.

Example: If the detection point is set to "2020-12-27T00:00:00Z", the granularity to "3600s" and the horizonDuration to "10800s", then we will generate 3 time series points (from "2020-12-27T01:00:00Z" to "2020-12-27T04:00:00Z"), for which we will return their forecasted values.

Note: The horizon time is only used for forecasting not for anormaly detection. To detect anomalies for multiple points of time, simply send multiple queries with those as detectionTime.

A duration in seconds with up to nine fractional digits, ending with 's'. Example: "3.5s".

noiseThreshold

number

Optional. Penalize variations between the actual and forecasted values smaller than this. For more information about how this parameter affects the score, see the anomalyScore formula.

Intuitively, anomaly scores summarize how statistically significant the change between the actual and forecasted value is compared with what we expect the change to be (see expectedDeviation). However, in practice, depending on the application, changes smaller than certain absolute values, while statistically significant, may not be important.

This parameter allows us to penalize such low absolute value changes.

Must be in the (0.0, inf) range.

If unspecified, it defaults to 0.000001.

Period

A time period of a fixed interval.

Enums
`PERIOD_UNSPECIFIED`	Unknown or simply not given.
`HOURLY`	1 hour
`DAILY`	24 hours
`WEEKLY`	7 days
`MONTHLY`	30 days
`YEARLY`	365 days