Data Clustering via Principal Direction Gap Partitioning

TL;DR

This paper analyzes the PCA-based clustering algorithm PDDP, identifies its limitations, and proposes a new gap partitioning method that leverages natural data gaps for improved clustering.

Contribution

The paper introduces a novel gap partitioning approach that enhances PCA-based clustering by considering natural data gaps, addressing previous method limitations.

Findings

The new method is comparable to PDDP on standard datasets.

Identifies specific scenarios where PDDP fails.

Provides geometric insights into PCA space for clustering.

Abstract

We explore the geometrical interpretation of the PCA based clustering algorithm Principal Direction Divisive Partitioning (PDDP). We give several examples where this algorithm breaks down, and suggest a new method, gap partitioning, which takes into account natural gaps in the data between clusters. Geometric features of the PCA space are derived and illustrated and experimental results are given which show our method is comparable on the datasets used in the original paper on PDDP.