東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Contributions to classification with...

Li, Li.

Linked to FindBook

Google Book

Amazon

博客來

Contributions to classification with missing values and variable selection in clustering.

Record Type:	Language materials, printed : Monograph/item
Title/Author:	Contributions to classification with missing values and variable selection in clustering./
Author:	Li, Li.
Description:	115 p.
Notes:	Adviser: James E. Gentle.
Contained By:	Dissertation Abstracts International68-12B.
Subject:	Statistics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3294022
ISBN:	9780549381662

Contributions to classification with missing values and variable selection in clustering.
Li, Li.

Contributions to classification with missing values and variable selection in clustering. - 115 p.

Adviser: James E. Gentle.

Thesis (Ph.D.)--George Mason University, 2008.

Meta learning plays an important role in selecting classifiers. It builds connections between data characteristics and classification performance. The pattern of missing values is an important data characteristic in classification problems. I investigated the relationship between missing values and the performance of tree-based classification methods. The subject area is another important factor in meta learning since datasets from the same subject area share common characteristics. I focused attention on data of land cover mapping related to invasive species. The land cover datasets are classified using tree-based classification methods. My study shows that random forests classifier has a better performance in terms of classification accuracy than single tree methods and linear separation. Researchers have studied clusters defining by different regression models and use this to identify and explain the natural structure of data. The widely used maximum likelihood approach includes all covariates in all clusters, which makes it difficult to see clusters or other interesting trends in the data. I proposed a clustering method that integrates variable selection using the lasso regression (Tibshirani, 1996). The new method can simultaneously partition data into groups and find effective covariates within each cluster. I conducted Monte Carlo studies to illustrate and verify the new method.

ISBN: 9780549381662Subjects--Topical Terms:

517247
Statistics.

Contributions to classification with missing values and variable selection in clustering.
LDR:02281nam 2200265 a 45 001 949339
005 20110525
008 110525s2008 ||||||||||||||||| ||eng d
020 $a 9780549381662
035 $a (UMI)AAI3294022
035 $a AAI3294022
040 $a UMI $c UMI
100 1 $a Li, Li. $3 1036040
245 1 0 $a Contributions to classification with missing values and variable selection in clustering.
300 $a 115 p.
500 $a Adviser: James E. Gentle.
500 $a Source: Dissertation Abstracts International, Volume: 68-12, Section: B, page: 8111.
502 $a Thesis (Ph.D.)--George Mason University, 2008.
520 $a Meta learning plays an important role in selecting classifiers. It builds connections between data characteristics and classification performance. The pattern of missing values is an important data characteristic in classification problems. I investigated the relationship between missing values and the performance of tree-based classification methods. The subject area is another important factor in meta learning since datasets from the same subject area share common characteristics. I focused attention on data of land cover mapping related to invasive species. The land cover datasets are classified using tree-based classification methods. My study shows that random forests classifier has a better performance in terms of classification accuracy than single tree methods and linear separation. Researchers have studied clusters defining by different regression models and use this to identify and explain the natural structure of data. The widely used maximum likelihood approach includes all covariates in all clusters, which makes it difficult to see clusters or other interesting trends in the data. I proposed a clustering method that integrates variable selection using the lasso regression (Tibshirani, 1996). The new method can simultaneously partition data into groups and find effective covariates within each cluster. I conducted Monte Carlo studies to illustrate and verify the new method.
590 $a School code: 0883.
650 4 $a Statistics. $3 517247
690 $a 0463
710 2 $a George Mason University. $3 1019450
773 0 $t Dissertation Abstracts International $g 68-12B.
790 $a 0883
790 1 0 $a Gentle, James E., $e advisor
791 $a Ph.D.
792 $a 2008
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3294022