import { Dataset } from './common';
export interface MakeClassificationProps {
    /** total number of samples */
    nSamples?: number;
    /** total number of features */
    nFeatures?: number;
    /** number of informative features */
    nInformative?: number;
    /** number of redundant features (random linear combinations of the informative ones) */
    nRedundant?: number;
    /** number of duplicated features (copies of informative/redundant columns) */
    nRepeated?: number;
    /** number of classes */
    nClasses?: number;
    /** number of gaussian clusters per class */
    nClustersPerClass?: number;
    /**
     * class proportions; length nClasses, or nClasses - 1 (the last weight
     * is inferred as 1 - sum)
     */
    weights?: number[];
    /** fraction of samples whose class is assigned randomly (label noise) */
    flipY?: number;
    /** factor multiplying the hypercube size; larger values spread the classes apart */
    classSep?: number;
    /** put cluster centroids on the vertices of a hypercube (vs a random polytope) */
    hypercube?: boolean;
    /**
     * shift added to all features; a single value or one per feature.
     * Pass null to draw a random shift in [-classSep, classSep] (sklearn's
     * `shift=None` behaviour).
     */
    shift?: number | number[] | null;
    /**
     * factor multiplying all features; a single value or one per feature.
     * Pass null to draw a random scale in [1, 100] (sklearn's `scale=None`
     * behaviour). Scaling happens after shifting.
     */
    scale?: number | number[] | null;
    /** shuffle the samples and the feature columns (default true) */
    shuffle?: boolean;
    /** seed for reproducible output */
    randomState?: number;
}
/**
 * Generates a random n-class classification problem
 * (port of sklearn.datasets.make_classification).
 *
 * Informative features are drawn from gaussian clusters placed on the
 * vertices of an nInformative-dimensional hypercube with sides of length
 * 2 * classSep, each cluster given a random covariance. Redundant features
 * are random linear combinations of the informative ones, repeated
 * features are exact duplicates, and the remaining features are noise.
 */
export declare function makeClassification(props?: MakeClassificationProps): Dataset;