東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Data fusion: A first step in decisio...

Hu, Jiaqi.

FindBook

Google Book

Amazon

博客來

Data fusion: A first step in decision informatics.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Data fusion: A first step in decision informatics./
作者:	Hu, Jiaqi.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2008,
面頁冊數:	89 p.
附註:	Source: Dissertation Abstracts International, Volume: 70-05, Section: B, page: 3150.
Contained By:	Dissertation Abstracts International70-05B.
標題:	Systems science. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3357222
ISBN:	9781109141832

Data fusion: A first step in decision informatics.
Hu, Jiaqi.

Data fusion: A first step in decision informatics. - Ann Arbor : ProQuest Dissertations & Theses, 2008 - 89 p.

Source: Dissertation Abstracts International, Volume: 70-05, Section: B, page: 3150.

Thesis (Ph.D.)--Rensselaer Polytechnic Institute, 2008.

This research proposes to develop data fusion tools that allow for the simultaneous utilization of qualitative and quantitative data in clustering analysis. From the data fusion perspective, the integration of multiple data sources can happen at different levels including data level, feature level, similarity level and decision level. We categorize methods that have been employed in the simultaneous utilization of qualitative and quantitative data, based on the fusion levels. We highlight two critical research areas where multiple data sources are to be fused at the feature and similarity level, respectively, which implies less information loss and potential flexible integration scheme. For the feature level fusion, we extend a probabilistic model for the mixed type data modeling to model the dependency between the qualitative and quantitative data, and embed the feature identification in the model estimation procedure. We also propose a model initialization strategy to reduce the influence of the initial configuration on the model estimation. We formulate the model estimation in an optimization framework where the penalized log likelihood is maximized. For the similarity level fusion, we propose a sub-sampling based method to search for the weight configuration in the weight sum rule. We also propose a voting based threshold strategy for noise reduction when the max rule is applied. We show through empirical studies that fusing quantitative and qualitative data can produce better results in clustering analysis than using individual data sources alone.

ISBN: 9781109141832Subjects--Topical Terms:

3168411
Systems science.

Data fusion: A first step in decision informatics.
LDR:02477nmm a2200277 4500 001 2161491
005 20180907134546.5
008 190424s2008 ||||||||||||||||| ||eng d
020 $a 9781109141832
035 $a (MiAaPQ)AAI3357222
035 $a AAI3357222
040 $a MiAaPQ $c MiAaPQ
100 1 $a Hu, Jiaqi. $3 3349449
245 1 0 $a Data fusion: A first step in decision informatics.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2008
300 $a 89 p.
500 $a Source: Dissertation Abstracts International, Volume: 70-05, Section: B, page: 3150.
500 $a Advisers: James M. Tien; Wai Kin (Victor) Chan.
502 $a Thesis (Ph.D.)--Rensselaer Polytechnic Institute, 2008.
520 $a This research proposes to develop data fusion tools that allow for the simultaneous utilization of qualitative and quantitative data in clustering analysis. From the data fusion perspective, the integration of multiple data sources can happen at different levels including data level, feature level, similarity level and decision level. We categorize methods that have been employed in the simultaneous utilization of qualitative and quantitative data, based on the fusion levels. We highlight two critical research areas where multiple data sources are to be fused at the feature and similarity level, respectively, which implies less information loss and potential flexible integration scheme. For the feature level fusion, we extend a probabilistic model for the mixed type data modeling to model the dependency between the qualitative and quantitative data, and embed the feature identification in the model estimation procedure. We also propose a model initialization strategy to reduce the influence of the initial configuration on the model estimation. We formulate the model estimation in an optimization framework where the penalized log likelihood is maximized. For the similarity level fusion, we propose a sub-sampling based method to search for the weight configuration in the weight sum rule. We also propose a voting based threshold strategy for noise reduction when the max rule is applied. We show through empirical studies that fusing quantitative and qualitative data can produce better results in clustering analysis than using individual data sources alone.
590 $a School code: 0185.
650 4 $a Systems science. $3 3168411
690 $a 0790
710 2 $a Rensselaer Polytechnic Institute. $3 1019062
773 0 $t Dissertation Abstracts International $g 70-05B.
790 $a 0185
791 $a Ph.D.
792 $a 2008
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3357222