Package: genieclust 1.1.6.9003

genieclust: Fast and Robust Hierarchical Clustering with Noise Points Detection

A retake on the Genie algorithm (Gagolewski, 2021 <doi:10.1016/j.softx.2021.100722>), which is a robust hierarchical clustering method (Gagolewski, Bartoszuk, Cena, 2016 <doi:10.1016/j.ins.2016.05.003>). It is now faster and more memory efficient; determining the whole cluster hierarchy for datasets of 10M points in low dimensional Euclidean spaces or 100K points in high-dimensional ones takes only a minute or so. Allows clustering with respect to mutual reachability distances so that it can act as a noise point detector or a robustified version of 'HDBSCAN*' (that is able to detect a predefined number of clusters and hence it does not dependent on the somewhat fragile 'eps' parameter). The package also features an implementation of inequality indices (e.g., Gini and Bonferroni), external cluster validity measures (e.g., the normalised clustering accuracy, the adjusted Rand index, the Fowlkes-Mallows index, and normalised mutual information), and internal cluster validity indices (e.g., the Calinski-Harabasz, Davies-Bouldin, Ball-Hall, Silhouette, and generalised Dunn indices). See also the 'Python' version of 'genieclust' available on 'PyPI', which supports sparse data, more metrics, and even larger datasets.

Authors:Marek Gagolewski [aut, cre, cph], Maciej Bartoszuk [ctb], Anna Cena [ctb], Peter M. Larsen [ctb]