Biclustering: Methods, Software and Application [Elektronische Ressource] / Sebastian Kaiser. Betreuer: Friedrich Leisch

Sebastian Kaiser

Informations

• Publié par: ludwig-maximilians-universitat_munchen
• Publié le: 01 janvier 2011
• Nombre de lectures: 36
• Langue: Deutsch
• Poids de l'ouvrage: 6 Mo

Extrait

Biclustering: Methods, Software and
Application
Sebastian Kaiser
Munchen 2011Biclustering: Methods, Software and
Application
Dissertation
zur Erlangung des akademischen Grades
eines Doktors der Naturwissenschaften am Institut fur Statistik
an der Fakultat fur Mathematik, Informatik und Statistik
der Ludwig-Maximilians-Universitat Munchen
Vorgelegt von
Sebastian Kaiser
am 7. Marz 2011
in MunchenErstgutachter: Prof. Dr. Friedrich Leisch, LMU Munchen
Zweitgutachter: Prof. Dr. Luis A. M. Quintales, Universidad de Salamanca
Rigorosum: 12. Mai 2011Research: the nal frontier. These are
the voyages of a young research fellow.
His ve-year mission: to explore
strange new methods, to seek out new
software and new applications, to
boldly go where no man has gone
before.
Based on Star TrekDanksagung
Danken mochte ich:
... meinem Doktorvater Fritz Leisch fur das Ermoglichen dieser Dissertation, fur
die vielen Freiheiten uber die letzten Jahre, fur die zahlreichen Mogl ichkeiten, das
wissenschaftliche Leben kennen zu lernen und die immer o ene T ur.
... den weiteren Gutachtern und Prufern Luis Quintales, Helmut Kuchenho , Chris-
tian Heumann und Volker Schmid fur ihr Interesse und ihre Zeit.
... meinen Kollegen am Institut fur Statistik fur die angenehme Arbeitsatmosphare
und das gute Klima, insbesondere Manuel Eugster, der mir das Arbeiten zum
Vergnugen hat werden lassen, und Carolin Strobl, die immer wieder fur Abwech-
slung im Buro gesorgt hat.
... der Visualisierungsgruppe aus Salamanca, insbesondere Rodrigo Santamaria fur
die gro artige Zusammenarbeit am biclust Paket.
... der Arbeitsgruppe aus Hasselt und Martin Sill fur die Zusammenarbeit an den
Bicluster Projekten.
... Sara Dolnicar und ihrer Arbeitsgruppe in Australien fur den wunderschonen
Aufenthalt und die tolle Zusammenarbeit.
... meinen Eltern, meiner Schwester Anna, meiner Oma Helga und meiner restlichen
Familie fur den Ruckhalt und das schon e Leben.
... meiner Freundin Maria, moge es immer so perfekt bleiben wie es ist.
... Fabian Barth fur die Mittagessen am Freitag und den Kontakt zur RSU - Suit
up!Abstract
Over the past 10 years, biclustering has become popular not only in the eld
of biological data analysis but also in other applications with high-dimensional
two way datasets. This technique clusters both rows and columns simulta-
neously, as opposed to clustering only rows or only columns. Biclustering
retrieves subgroups of objects that are similar in one subgroup of variables
and dierent in the remaining variables. This dissertation focuses on improv-
ing and advancing biclustering methods. Since most existing methods are
extremely sensitive to variations in parameters and data, we developed an en-
semble method to overcome these limitations. It is possible to retrieve more
stable and reliable bicluster in two ways: either by running algorithms with dif-
ferent parameter settings or by running them on sub- or bootstrap samples of
the data and combining the results. To this end, we designed a software pack-
age containing a collection of bicluster algorithms for di erent clustering tasks
and data scales, developed several new ways of visualizing bicluster solutions,
and adapted traditional cluster validation indices (e.g. Jaccard index) for val-
idating the bicluster framework. Finally, we applied biclustering to marketing
data. Well-established algorithms were adjusted to slightly di erent data sit-
uations, and a new method specially adapted to ordinal data was developed.
In order to test this method on articial data, we generated correlated original
random values. This dissertation introduces two methods for generating such
values given a probability vector and a correlation structure.
All the methods outlined in this dissertation are freely available in the R pack-
ages biclust and orddata. Numerous examples in this work illustrate how to
use the methods and software.