Online hierarchical partitioning of the output space in extreme multi-label data stream

TL;DR

This paper introduces iHOMER, an online hierarchical clustering framework for multi-label data streams that dynamically adapts to concept drift, outperforming existing methods in accuracy and robustness.

Contribution

iHOMER is the first online method to incrementally partition label space into correlated clusters without predefined hierarchies, using online clustering and drift detection.

Findings

iHOMER outperforms 5 state-of-the-art global baselines by 23%.

iHOMER outperforms 12 local baselines by 32%.

Demonstrates robustness across 23 real-world datasets.

Abstract

Mining data streams with multi-label outputs poses significant challenges due to evolving distributions, high-dimensional label spaces, sparse label occurrences, and complex label dependencies. Moreover, concept drift affects not only input distributions but also label correlations and imbalance ratios over time, complicating model adaptation. To address these challenges, structured learners are categorized into local and global methods. Local methods break down the task into simpler components, while global methods adapt the algorithm to the full output space, potentially yielding better predictions by exploiting label correlations. This work introduces iHOMER (Incremental Hierarchy Of Multi-label Classifiers), an online multi-label learning framework that incrementally partitions the label space into disjoint, correlated clusters without relying on predefined hierarchies. iHOMER leverages online divisive-agglomerative clustering based on \textit{Jaccard} similarity and a global tree-based learner driven by a multivariate \textit{Bernoulli} process to guide instance partitioning. To address non-stationarity, it integrates drift detection mechanisms at both global and local levels, enabling dynamic restructuring of label partitions and subtrees. Experiments across 23 real-world datasets show iHOMER outperforms 5 state-of-the-art global baselines, such as MLHAT, MLHT of Pruned Sets and iSOUPT, by 23\%, and 12 local baselines, such as binary relevance transformations of kNN, EFDT, ARF, and ADWIN bagging/boosting ensembles, by 32\%, establishing its robustness for online multi-label classification.