Class Dominancy Profiles in Multi-class and Multi-cluster Similarity Networks

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00098892%3A_____%2F25%3A10159372" target="_blank" >RIV/00098892:_____/25:10159372 - isvavai.cz</a>

Nalezeny alternativní kódy

RIV/61989100:27240/25:10257850

Výsledek na webu

<a href="https://link.springer.com/chapter/10.1007/978-3-031-82439-5_2" target="_blank" >https://link.springer.com/chapter/10.1007/978-3-031-82439-5_2</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-3-031-82439-5_2" target="_blank" >10.1007/978-3-031-82439-5_2</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Class Dominancy Profiles in Multi-class and Multi-cluster Similarity Networks

Popis výsledku v původním jazyce

The visual representation of data analysis results, a current trend in data science, is essential for people without deep knowledge of statistics and machine learning. Patient Similarity Networks (PSNs), as an instance of similarity networks in general, are commonly used visualization tools in biomedical data analysis. PSNs are an understandable presentation of the complex relationships hidden in patient data, such as partitioning patients with similar characteristics into clusters, providing relatively easy interpretation to clinicians and clinical biologists. However, the interpretation may be inaccurate due to an incorrect intuition, especially in complex multi-class, multi-cluster situations. Our paper focuses on analyzing cluster-class relationships in PSNs, assuming different numbers of classes and clusters, both of different and potentially imbalanced sizes. We use several problematic situations to show how to support clinical intuition in interpreting these situations correctly. We use the Matthews correlation coefficient to analyze cluster-class relationships in a way that understandably complements the PSN visualization. Finally, we present a visualization of this approach on a real-world patient similarity network.

Název v anglickém jazyce

Class Dominancy Profiles in Multi-class and Multi-cluster Similarity Networks

Popis výsledku anglicky

The visual representation of data analysis results, a current trend in data science, is essential for people without deep knowledge of statistics and machine learning. Patient Similarity Networks (PSNs), as an instance of similarity networks in general, are commonly used visualization tools in biomedical data analysis. PSNs are an understandable presentation of the complex relationships hidden in patient data, such as partitioning patients with similar characteristics into clusters, providing relatively easy interpretation to clinicians and clinical biologists. However, the interpretation may be inaccurate due to an incorrect intuition, especially in complex multi-class, multi-cluster situations. Our paper focuses on analyzing cluster-class relationships in PSNs, assuming different numbers of classes and clusters, both of different and potentially imbalanced sizes. We use several problematic situations to show how to support clinical intuition in interpreting these situations correctly. We use the Matthews correlation coefficient to analyze cluster-class relationships in a way that understandably complements the PSN visualization. Finally, we present a visualization of this approach on a real-world patient similarity network.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2025

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Complex Networks &amp; Their Applications XIII

ISBN

978-3-031-82438-8

ISSN

1860-949X

e-ISSN

1860-9503

Počet stran výsledku

12

Strana od-do

15-26

Název nakladatele

Springer Cham

Místo vydání

Cham

Místo konání akce

Istanbul

Datum konání akce

10. 12. 2024

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

001489045000002