High-Dimensional Data Mining: Subspace Clustering, Outlier Detection and applications to classification - Couverture souple

Foss, Andrew

 
9783639362114: High-Dimensional Data Mining: Subspace Clustering, Outlier Detection and applications to classification
Couverture souple
ISBN 10 : 363936211X ISBN 13 : 9783639362114
Editeur : VDM Verlag Dr. Müller, 2011

Synopsis

Data mining in high dimensionality typically faces the consequences of increasing sparsity and declining differentiation between points, while sparsity tends to increase false negatives. Here, the problem of solving high-dimensional problems using low-dimensional solutions is addressed. In clustering, we provide a new framework for finding candidate subspaces and the clusters within them using only two-dimensional clustering. It is robust to noise and handles overlapping clusters. In the field of outlier detection, several novel algorithms suited to high-dimensional data are presented.m These outperform state-of-the-art outlier detection algorithms in ranking outlierness for many datasets regardless of whether they contain rare classes or not. This approach can be a powerful means of classification for heavily overlapping classes given sufficiently high dimensionality. This is achieved solely due to the differences in variance among the classes. On some difficult datasets, this unsupervised approach yielded better separation than the very best supervised classifiers. This opens a new field in data mining, classification through differences in variance rather than spatial location.

Les informations fournies dans la section « Synopsis » peuvent faire référence à une autre édition de ce titre.

Présentation de l'éditeur

Data mining in high dimensionality typically faces the consequences of increasing sparsity and declining differentiation between points, while sparsity tends to increase false negatives. Here, the problem of solving high-dimensional problems using low-dimensional solutions is addressed. In clustering, we provide a new framework for finding candidate subspaces and the clusters within them using only two-dimensional clustering. It is robust to noise and handles overlapping clusters. In the field of outlier detection, several novel algorithms suited to high-dimensional data are presented.m These outperform state-of-the-art outlier detection algorithms in ranking outlierness for many datasets regardless of whether they contain rare classes or not. This approach can be a powerful means of classification for heavily overlapping classes given sufficiently high dimensionality. This is achieved solely due to the differences in variance among the classes. On some difficult datasets, this unsupervised approach yielded better separation than the very best supervised classifiers. This opens a new field in data mining, classification through differences in variance rather than spatial location.

Biographie de l'auteur

Andrew Foss received his BA in Physics and MA from St John's College, Oxford and an MSc and PhD in Computing Science from the University of Alberta. He has worked extensively in IT, developing software that sells internationally. His particular interest is in forcasting.

Les informations fournies dans la section « A propos du livre » peuvent faire référence à une autre édition de ce titre.

Éditeur
VDM Verlag Dr. Müller
Date d'édition
2011
Langue
anglais
ISBN 10 
363936211X
ISBN 13 
9783639362114
Reliure
Broché
Nombre de pages
152
Coordonnées du fabricant
non disponible
Personne responsable
non disponible