On the strengths of the self-updating process clustering algorithm

TL;DR

The paper presents the Self-Updating Process (SUP), a flexible clustering algorithm that iteratively models data point movements, excelling in noisy, large, and unbalanced datasets, with demonstrated effectiveness through simulations and real data.

Contribution

It introduces SUP, a novel, generalizable clustering algorithm extending correlation matrix methods and capable of handling noise, large cluster numbers, and unbalanced data.

Findings

SUP effectively isolates noise points during clustering.

SUP performs well with large numbers of clusters.

SUP demonstrates strong results in simulations and real data applications.

Abstract

We introduce a simple, intuitive and yet powerful algorithm for clustering analysis. This algorithm is an iterative process on the sample space, which arises as an extension of the iteratively generated correlation matrices. It allows for both time-varying and time-invariant operators, therefore can be considered more general than the blurring mean-shift algorithm in which operators are time-invariant. The algorithm stands from the viewpoint of data points and simulates the process how data points move and perform self-clustering, therefore is named Self-Updating Process (SUP). It is particularly competitive for (i) data with noise, (ii) data with large number of clusters and (iii) unbalanced data. When noise is present in the data, the algorithm is able to isolate noisy points while performing clustering simultaneously. Simulation studies and real data applications are presented to demonstrate the performance of SUP.