interface Cluster {
    id: number;
    position: number;
    childs: number[];
}
/**
 * Simple 1D K-means clustering.
 * Runs multiple attempts and returns the best result (most balanced cluster sizes).
 * @param k Number of clusters (default 2).
 * @param arr Input data points.
 * @param attempts Number of attempts to run (default 1).
 */
declare const KMeans: (k: number | undefined, arr: number[], attempts?: number) => Cluster[];
interface NDCluster<T> {
    /** 1-based cluster id */
    id: number;
    /** Mean position in feature space */
    centroid: number[];
    /** Original data items assigned to this cluster */
    members: T[];
    /** Number of members */
    size: number;
    /**
     * Within-cluster sum of squares — sum of squared Euclidean distances
     * from each member to the centroid. Lower means a tighter, more
     * cohesive cluster.
     */
    wcss: number;
}
type FeatureExtractor<T> = (item: T) => number[];
interface KMeansNDOptions {
    /**
     * Number of independent runs. The run with the lowest total WCSS wins.
     * More attempts → more reliable result at the cost of compute.
     * Default: 5.
     */
    attempts?: number;
    /**
     * Centroid initialisation strategy.
     * - `'kmeans++'` (default) spreads initial centroids far apart,
     *   which dramatically reduces the chance of a poor local minimum.
     * - `'random'` picks k random data points as starting centroids.
     */
    init?: 'random' | 'kmeans++';
    /**
     * Hard cap on iteration count per attempt.
     * Default: 300.
     */
    maxIter?: number;
}
/**
 * N-dimensional K-means clustering suitable for production analytics
 * (customer segmentation, product segmentation, etc.).
 *
 * @param k Number of clusters.
 * @param data Array of items to cluster.
 * @param features Function that extracts a numeric feature vector from each item.
 *   All vectors must have the same length.
 * @param options Optional tuning parameters.
 *
 * @example
 * // Customer segmentation by recency, frequency, monetary value (RFM)
 * const segments = KMeansND(3, customers, (c) => [c.recency, c.frequency, c.spend]);
 *
 * @example
 * // Works with plain number arrays too
 * const result = KMeansND(2, [[1,2],[3,4],[100,200]], (x) => x);
 */
declare const KMeansND: <T>(k: number, data: T[], features: FeatureExtractor<T>, options?: KMeansNDOptions) => NDCluster<T>[];

export { type FeatureExtractor, KMeans, KMeansND, type KMeansNDOptions, type NDCluster };