TY - GEN

T1 - Comparative Study of Clustering Algorithms for Inferring Psychological Profiles from VK-User Avatars Semantics

AU - Bushmelev, F.

AU - Stoliarova, V.

AU - Prusskikh , Ilya

N1 - Export Date: 29 March 2026; Cited By: 0; Correspondence Address: F. Bushmelev; St. Petersburg Federal Research Center of the Russian Academy of Sciences, St. Petersburg, 39, 14th Line V.O., 199178, Russian Federation; email: fvb@dscs.pro; Conference name: 9th International Scientific Conference on Intelligent Information Technologies for Industry, IITI 2025; Conference date: 5 November 2025 through 7 November 2025; Conference code: 344719

PY - 2026

Y1 - 2026

N2 - Rising user activity on online social media (OSM) platforms like VK drives cross-disciplinary research (psychology, cybersecurity, etc.), where avatars serve as key digital footprints. While ML is widely used for the analysis, adapting universal tools to dataset-specific properties remains challenging. This study focuses on the optimization of the clustering of the datasets with avatars. The intensive computational experiment was conducted in order to identify the clustering structure of such dataset and the best UMAP parameter values that lead to good clusterization with respect to several clusterization quality indices. Our pipeline combines CLIP embeddings, UMAP reduction, and five clustering algorithms (K-means to HDBSCAN and GMM). Hyperparameters were tuned via Grid Search and Bayesian optimization, evaluated on 9,000 VK avatars using four metrics (SI, DBI, CHI, DI). We demonstrate, that those parameters lead to avatar clusterization with user groups that vary in mean Big Five scales. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

AB - Rising user activity on online social media (OSM) platforms like VK drives cross-disciplinary research (psychology, cybersecurity, etc.), where avatars serve as key digital footprints. While ML is widely used for the analysis, adapting universal tools to dataset-specific properties remains challenging. This study focuses on the optimization of the clustering of the datasets with avatars. The intensive computational experiment was conducted in order to identify the clustering structure of such dataset and the best UMAP parameter values that lead to good clusterization with respect to several clusterization quality indices. Our pipeline combines CLIP embeddings, UMAP reduction, and five clustering algorithms (K-means to HDBSCAN and GMM). Hyperparameters were tuned via Grid Search and Bayesian optimization, evaluated on 9,000 VK avatars using four metrics (SI, DBI, CHI, DI). We demonstrate, that those parameters lead to avatar clusterization with user groups that vary in mean Big Five scales. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

KW - CLIP

KW - Clustering

KW - Dimensionality Reduction

KW - Graphical Digital Footprints

KW - Hyperparameter Tuning

KW - Online Social Media

KW - Personality Computing

KW - Barium compounds

KW - Bayesian networks

KW - Computer graphics

KW - Dimensionality reduction

KW - K-means clustering

KW - Reduction

KW - Social sciences computing

KW - Clusterings

KW - Clusterization

KW - Comparatives studies

KW - Graphical digital footprint

KW - Hyper-parameter

KW - Hyperparameter tuning

KW - Online social medias

KW - Personality computing

KW - Social networking (online)

UR - https://www.mendeley.com/catalogue/807b4a9b-b6ce-31a8-b152-3c9751bb824b/

U2 - 10.1007/978-3-032-13615-2_41

DO - 10.1007/978-3-032-13615-2_41

M3 - статья в сборнике материалов конференции

SN - 9783032136145

T3 - Lecture Notes in Networks and Systems

SP - 485

EP - 496

BT - Proceedings of the Ninth International Scientific Conference “Intelligent Information Technologies for Industry” (IITI’25), Volume 1

PB - Springer Nature

Y2 - 5 November 2025 through 7 November 2025

ER -