FLASC: a flare-sensitive clustering algorithm
Daniël M. Bot, Jannes Peeters, Jori Liesenborgs, Jan A Aerts · PeerJ Computer Science · 2025
Exploratory data analysis workflows often use clustering algorithms to find groups of similar data points. The shape of these clusters can provide meaningful information about the data. For example, a Y-shaped cluster might represent an evolving process with two distinct outcomes. This article presents flare-sensitive clustering (FLASC), an algorithm that detects branches within clusters to identify such shape-based subgroups. FLASC builds upon HDBSCAN*—a state-of-the-art density-based clustering algorithm—and detects branches in a post-processing step using within-cluster connectivity. Two algorithm variants are presented, which trade computational cost for noise robustness. We show that both variants scale similarly to HDBSCAN* regarding computational cost and provide similar outputs across repeated runs. In addition, we demonstrate the benefit of branch detection on two real-world data sets. Our implementation is included in the hdbscan Python package and available as a standalone package at https://github.com/vda-lab/pyflasc .