Your browser does not allow JavaScript!
JavaScript is necessary for the proper functioning of this website. Please enable JavaScript or use a modern browser.
Open Science Slovenia
Open Science
DiKUL
slv
|
eng
Search
Browse
New in RUL
About RUL
In numbers
Help
Sign in
Označevanje skupin primerov v točkovnih prikazih podatkov
ID
Janežič, Simon
(
Author
),
ID
Zupan, Blaž
(
Mentor
)
More about this mentor...
PDF - Presentation file,
Download
(3,60 MB)
MD5: 192D088A82FF8C067C28DF916742B70E
Image galllery
Abstract
Za boljše razumevanje visoko dimenzionalnih podatkov analitiki pogosto uporabljamo projekcije ali uvrščanje podatkov v nižje dimenzije. Ti pristopi omogočajo izris točkovnega prikaza podatkov v dveh dimenzijah, ki ohranja strukturo prvotnih podatkov. Interpretacija teh prikazov zahteva dodatno ročno analizo. V našem delu predlagamo pristop za strojno razlago dvodimenzionalnih predstavitev podatkov. Predlagana metoda temelji na pristopu za razlago dvodimenzionalnih predstavitev besedilnih dokumentov, ki ga v našem delu razširimo za uporabo na splošnih tabelaričnih podatkih. Pristop na točkovnih prikazih poišče skupine točk in pripadajoče oznake, ki primere v skupinah opišejo v jeziku atributov, s katerimi so bili podatki prvotno opisani. Skupine točk določimo z algoritmom za razvrščanje DBSCAN. Značilne oznake skupin pridobimo s pomočjo statističnih testov. Metoda omogoča tudi interaktivno raziskovanje in označevanje poljubnih podskupin, ki se uporabniku zdijo zanimive. Uporabnost pristopa prikažemo na analizi različnih podatkovnih množic. Pokažemo, da je metoda koristna za učinkovito analizo dvodimenzionalnih predstavitev podatkov in prikaz ključnih značilnosti podatkov.
Language:
Slovenian
Keywords:
točkovni prikazi
,
razsevni diagram
,
označevanje skupin
,
razlaga razvrščanja v skupine
,
gručenje
Work type:
Master's thesis/paper
Organization:
FRI - Faculty of Computer and Information Science
Year:
2019
PID:
20.500.12556/RUL-111856
COBISS.SI-ID:
1538419395
Publication date in RUL:
16.10.2019
Views:
1227
Downloads:
240
Metadata:
Cite this work
Plain text
BibTeX
EndNote XML
EndNote/Refer
RIS
ABNT
ACM Ref
AMA
APA
Chicago 17th Author-Date
Harvard
IEEE
ISO 690
MLA
Vancouver
:
Copy citation
Share:
Secondary language
Language:
English
Title:
Label inference for data clusters in point-based visualizations
Abstract:
Two-dimensional point-based visualizations of multidimensional data may reveal data structures and clusters that require further interpretation. We present an approach that can automatically annotate the clusters in these visualisations. Our method extends the existing procedure for automatic annotation of two-dimensional representations of text documents and enables it for general attribute-value data. We propose to finds groups of points on scatterplot visualisations and assign them labels that describe a group’s characteristics in a language of the attributes of the original data. The approach uses DBSCAN clustering algorithm to find groups of points in the scatterplots. Statistical tests are used to determine labels for each of the groups. The proposed approach also features an interactive exploration of arbitrary subgroups manually chosen by the user. We analyze three datasets to demonstrate the usefulness of our approach. We show that the proposed method is sufficiently fast to support interactive analysis and that the group annotations found by our approach are meaningful.
Keywords:
scatterplots
,
cluster annotation
,
explanation of clustering results
,
clustering
Similar documents
Similar works from RUL:
Similar works from other Slovenian collections:
Back