Comparative Validation of Karate Kumite Performance Profiles Using Multiple Clustering Methods

This study examines when clusters derived from karate kumite match data are sufficiently corroborated to be interpreted as athlete-performance profiles.Data comprised 114 athlete-match observations from 57 female WKF matches.Rather than relying on cluster separation (silhouette) from a single algorithm, as commonly reported, five algorithms with differing assumptions (Ward, k-Means, DBSCAN, RSKC-STOF, ARSK) clustered the data using outcome-independent features.These clusters were considered corroborated only if consistent across algorithms, stronger than a permutation null, not representation artifacts, and reproducible under match and athlete resampling with model re-selection.Methodologically, comparison across algorithms, representation analysis, and repeated model selection each demonstrated the insufficiency of cluster separation from a single algorithm.All silhouettes were weak (0.31-0.49), and comparing algorithms showed the five Female-47 kg clusters from k-Means alone disagreed with other algorithms (ARI = 0.24-0.58),indicating algorithm-dependent clusters; representation analysis showed the Female-54 kg partition, though algorithm-independent, merely separated scorers from non-scorers (ARI = 1.0), indicating a representation artifact; and repeated model selection showed the Female-42 kg four-cluster solution recurred in 29-38% of resamples, leaving the cluster number uncertain.For karate performance analysis, the clusters mostly reflected scoring rather than fighting behavior: attack and defend did not differ between Female-54 kg or main Female-42 kg clusters (overlapping 95% CIs).For the present data, only Female-42 kg contained partially corroborated profiles, non-scorers (n = 14) and high-yuko scorers (n = 15; yuko = 4.67), whereas Female-47 kg was not corroborated and Female-54 kg was reproducible but not corroborated.

Authors

Institutions

Publication Details

Journal
International journal of intelligent engineering and systems
Published
2026-11-17
DOI
https://doi.org/10.22266/ijies2026.1231.73
Primary Topic
Martial Arts: Techniques, Psychology, and Education
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Comparative Validation of Karate Kumite Performance Profiles Using Multiple Clustering Methods

Derry Alamsyah, Fajar Awang Irawan, Dewangga Yudhistira, Samsuryadi et al.
International journal of intelligent engineering and systems
Martial Arts: Techniques, Psychology, and Education
article

Comparative Validation of Karate Kumite Performance Profiles Using Multiple Clustering Methods

Derry Alamsyah, Fajar Awang Irawan, Dewangga Yudhistira, Samsuryadi, Hartati
article en

Abstract

This study examines when clusters derived from karate kumite match data are sufficiently corroborated to be interpreted as athlete-performance profiles.Data comprised 114 athlete-match observations from 57 female WKF matches.Rather than relying on cluster separation (silhouette) from a single algorithm, as commonly reported, five algorithms with differing assumptions (Ward, k-Means, DBSCAN, RSKC-STOF, ARSK) clustered the data using outcome-independent features.These clusters were considered corroborated only if consistent across algorithms, stronger than a permutation null, not representation artifacts, and reproducible under match and athlete resampling with model re-selection.Methodologically, comparison across algorithms, representation analysis, and repeated model selection each demonstrated the insufficiency of cluster separation from a single algorithm.All silhouettes were weak (0.31-0.49), and comparing algorithms showed the five Female-47 kg clusters from k-Means alone disagreed with other algorithms (ARI = 0.24-0.58),indicating algorithm-dependent clusters; representation analysis showed the Female-54 kg partition, though algorithm-independent, merely separated scorers from non-scorers (ARI = 1.0), indicating a representation artifact; and repeated model selection showed the Female-42 kg four-cluster solution recurred in 29-38% of resamples, leaving the cluster number uncertain.For karate performance analysis, the clusters mostly reflected scoring rather than fighting behavior: attack and defend did not differ between Female-54 kg or main Female-42 kg clusters (overlapping 95% CIs).For the present data, only Female-42 kg contained partially corroborated profiles, non-scorers (n = 14) and high-yuko scorers (n = 15; yuko = 4.67), whereas Female-47 kg was not corroborated and Female-54 kg was reproducible but not corroborated.

International journal of intelligent engineering and systemsVol. 19(12)
State University of Semarang (ID), Universitas Negeri Surabaya (ID), Sriwijaya University (ID), Universitas Multi Data Palembang (ID)
Ministrstvo za visoko šolstvo, znanost in tehnologijo, Lembaga Pengelola Dana Pendidikan
Martial Arts: Techniques, Psychology, and Education
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon