Single-pass Possibilistic Clustering with Damped Window Footprints

TL;DR

This paper introduces a single-pass possibilistic clustering algorithm designed for streaming data, capable of modeling complex cluster shapes and efficiently updating cluster footprints over damped windows, validated against multiple benchmarks.

Contribution

The paper presents a novel SPC algorithm that models non-spherical clusters, updates footprints over damped windows, and merges cluster estimates using covariance union, enhancing streaming clustering.

Findings

Outperforms five streaming clustering algorithms in purity and mutual information

Effectively models non-spherical clusters in streaming data

Provides closed-form updates for footprints over damped windows

Abstract

Streaming clustering is a domain that has become extremely relevant in the age of big data, such as in network traffic analysis or in processing continuously-running sensor data. Furthermore, possibilistic models offer unique benefits over approaches from the literature, especially with the introduction of a "fuzzifier" parameter that controls how quickly typicality degrades as one gets further from cluster centers. We propose a single-pass possibilistic clustering (SPC) algorithm that is effective and easy to apply to new datasets. Key contributions of SPC include the ability to model non-spherical clusters, closed-form footprint updates over arbitrarily sized damped windows, and the employment of covariance union from the multiple hypothesis tracking literature to merge two cluster mean and covariance estimates. SPC is validated against five other streaming clustering algorithm on the basis of cluster purity and normalized mutual information.