Chameleon2++: an efficient and scalable variant of chameleon clustering

Abstract

Hierarchical clustering remains a fundamental challenge in data mining, particularly when dealing with large-scale datasets where traditional approaches fail to scale e↵ectively. Recent Chameleon-based algorithms—Chameleon2 (2019), M-Chameleon (2021), and INNGS-Chameleon (2021)—have advanced strategies but su↵er from O(n2) computational complexity. Particularly in their graph generation stage due to exact k-NN computation. While tolerable on synthetic or small datasets, this quickly becomes a bottleneck for real-world datasets which are large-scale and high-dimensional. We introduce Chameleon2++, a scalable extension of Chameleon2 tailored for real-world applications.