Online clustering with interpretable drift adaptation to mixed features

Flavio Corradini, Vincenzo Nucci, Marco Piangerelli, Barbara Re

Abstract: Highlights•FURAKI: Unsupervised online clustering for mixed data with concept drift handling.•Drift Types: Detects abrupt, recurrent, incremental, and gradual concept drifts using KDE and G-tests.•Interpretable Model: Uses binary tree to show features driving changes in clustering.•Performance: Beats SOTA in F1-score and ARI on synthetic and real-world data.•Mixed Features: Handles numeric and categorical features without transformations.