Type Alias FeatureInsights

Metadata about the features in your dataset, generated when you create ProfileInsights.

type FeatureInsights = {
    cannotBeTarget: boolean;
    defaultFeatureType?: FeatureType;
    engineeredFeatures?: string[];
    estimatedMaxForecastHorizon?: number;
    experimentTypes: ExperimentType[];
    insights: Insights[];
    name: string;
    willBeDropped: boolean;
}
Index

Properties

cannotBeTarget defaultFeatureType? engineeredFeatures? estimatedMaxForecastHorizon? experimentTypes insights name willBeDropped

Properties

cannotBeTarget: boolean

Whether a feature cannot be the target field

defaultFeatureType?: FeatureType

The default feature type based on the feature's data type. If you want a value to be interpreted differently (e.g. 0/1 as categorical/boolean instead of numeric), use changeType.

engineeredFeatures?: string[]

Preliminary list of engineered features as strings. If subsequent processing validates them, they'll be converted to EngineeredFeature objects within a NestedColumn, each of which may contain its own FeatureInsights.

estimatedMaxForecastHorizon?: number

Only applies for time series experiment types. This intial estimate of the combined max forecast window and gap (aka - horizon). It only applies to possible date index columns. After the experiment version is created, we get a more precise number for subsequent versions. When training data is grouped, this estimate may be less accurate.

experimentTypes: ExperimentType[]

Experiment types in this feature insight

insights: Insights[]

List of insights about this feature.

name: string

Name of the feature insight

willBeDropped: boolean

Whether this feature will be dropped. Traits like high cardinality make some features less predictive or too costly to merit use.

Settings

Member Visibility

On This Page

Properties