Rethinking Cluster Quality Evaluation for Large-Scale Geospatial Data

Read full text Summarise with VIRAT
Authors: Maksym Arzubov, Olga Narushynska
Cited by: 0
References: 30

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper advances cluster quality evaluation for large-scale geospatial data by challenging the adequacy of classical validation metrics based on global or pairwise distance computations. Although widely used, such metrics fail to scale computationally and inadequately capture spatial heterogeneity, local density variation, and multi-scale spatial organization. We argue that cluster quality evaluation in this context requires a shift from global distance-based reasoning toward locality-driven assessment. To this end, we propose a scalable evaluation framework grounded in local spatial structure, which quantifies neighborhood-level compactness, density consistency, and spatial coherence. Local quality indicators are computed within spatial neighborhoods and aggregated into a global assessment without exhaustive pairwise analysis. The proposed framework is algorithm-agnostic and applicable to crisp, fuzzy, and hierarchical clustering results. Experimental validation on large real-world geospatial datasets demonstrates that the proposed approach delivers stable, interpretable, and scale-consistent quality estimates while significantly reducing computational complexity compared to conventional internal validation measures. Furthermore, the results reveal improved robustness to non-uniform spatial distributions and density gradients, exposing limitations of traditional metrics at global scales. Overall, the proposed framework constitutes a more appropriate evaluation paradigm for clustering large-scale geospatial data in modern analytical and interactive mapping environments.

Fields of Study & Subjects

Advanced Clustering Algorithms Research Data Management and Algorithms Advanced Decision-Making Techniques

Publication details

Journal / source
COMPUTER-INTEGRATED TECHNOLOGIES EDUCATION SCIENCE PRODUCTION
Issue
62
Pages
107–117
Publication date
28 March 2026
Language
EN