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

A Time Series Data Mining and Unobse...

Wilcox, Bruce A.

FindBook

Google Book

Amazon

博客來

A Time Series Data Mining and Unobserved Component Modeling Approach to Credit Risk Correlation Modeling.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	A Time Series Data Mining and Unobserved Component Modeling Approach to Credit Risk Correlation Modeling./
作者:	Wilcox, Bruce A.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2018,
面頁冊數:	305 p.
附註:	Source: Dissertations Abstracts International, Volume: 79-12, Section: B.
Contained By:	Dissertations Abstracts International79-12B.
標題:	Applied Mathematics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10789848
ISBN:	9780438017740

A Time Series Data Mining and Unobserved Component Modeling Approach to Credit Risk Correlation Modeling.
Wilcox, Bruce A.

A Time Series Data Mining and Unobserved Component Modeling Approach to Credit Risk Correlation Modeling. - Ann Arbor : ProQuest Dissertations & Theses, 2018 - 305 p.

Source: Dissertations Abstracts International, Volume: 79-12, Section: B.

Thesis (Ph.D.)--The Claremont Graduate University, 2018.

This item must not be added to any third party search indexes.

The objective of this thesis is to explore whether applying data mining methods from the fields of information systems and applied mathematics and state-space methods from the domain of systems engineering to the management of a portfolio of bonds or other financial obligations in a new way can provide new insights and tools for portfolio managers. Specifically, this thesis develops and applies time series clustering approaches with the intention of developing effective tools for diversification of bond portfolios. Traditional time series clustering techniques are based on specifying a relevant distance metric that captures the similarity or dissimilarity between any pair of time series and then executing an optimization routine to cluster time series in a way that maximizes inter-cluster distances while minimizing intra-cluster distances. Since the mid-1990s there has been significant research into the identification and validation of such distance-based approaches with varying degrees of success. This project defines and tests three such approaches. A recent and distinct approach to time series clustering known as model-based clustering is based on the specification and estimation of finite mixture models. This approach is based on fitting time series to generative models and the assignment of time series to these generative models. The approach of model-based clustering is to simultaneously fit both the generative model parameters and the allocations of elements to clusters to achieve the best probabilistic fit to the data observed. This thesis work represents the first known published definition and evaluation of model-based time series clustering using generative models that are specified in state space format. These four clustering approaches were developed in an extensive Matlab environment generated as part of this thesis and evaluated using techniques of synthetic bond data generation with a known ground truth and by examining the performance of portfolios constructed from an extensive database of actual bond transactions. The new model-based clustering approach was found to have tangible benefits both in providing portfolio diversification and intuitive insight into portfolio allocation decisions.

ISBN: 9780438017740Subjects--Topical Terms:

1669109
Applied Mathematics.

A Time Series Data Mining and Unobserved Component Modeling Approach to Credit Risk Correlation Modeling.
LDR:03436nmm a2200337 4500 001 2210880
005 20191126113850.5
008 201008s2018 ||||||||||||||||| ||eng d
020 $a 9780438017740
035 $a (MiAaPQ)AAI10789848
035 $a (MiAaPQ)cgu:11204
035 $a AAI10789848
040 $a MiAaPQ $c MiAaPQ
100 1 $a Wilcox, Bruce A. $3 2251420
245 1 0 $a A Time Series Data Mining and Unobserved Component Modeling Approach to Credit Risk Correlation Modeling.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2018
300 $a 305 p.
500 $a Source: Dissertations Abstracts International, Volume: 79-12, Section: B.
500 $a Publisher info.: Dissertation/Thesis.
500 $a Advisor: Hamano, Fumio.
502 $a Thesis (Ph.D.)--The Claremont Graduate University, 2018.
506 $a This item must not be added to any third party search indexes.
506 $a This item must not be sold to any third party vendors.
520 $a The objective of this thesis is to explore whether applying data mining methods from the fields of information systems and applied mathematics and state-space methods from the domain of systems engineering to the management of a portfolio of bonds or other financial obligations in a new way can provide new insights and tools for portfolio managers. Specifically, this thesis develops and applies time series clustering approaches with the intention of developing effective tools for diversification of bond portfolios. Traditional time series clustering techniques are based on specifying a relevant distance metric that captures the similarity or dissimilarity between any pair of time series and then executing an optimization routine to cluster time series in a way that maximizes inter-cluster distances while minimizing intra-cluster distances. Since the mid-1990s there has been significant research into the identification and validation of such distance-based approaches with varying degrees of success. This project defines and tests three such approaches. A recent and distinct approach to time series clustering known as model-based clustering is based on the specification and estimation of finite mixture models. This approach is based on fitting time series to generative models and the assignment of time series to these generative models. The approach of model-based clustering is to simultaneously fit both the generative model parameters and the allocations of elements to clusters to achieve the best probabilistic fit to the data observed. This thesis work represents the first known published definition and evaluation of model-based time series clustering using generative models that are specified in state space format. These four clustering approaches were developed in an extensive Matlab environment generated as part of this thesis and evaluated using techniques of synthetic bond data generation with a known ground truth and by examining the performance of portfolios constructed from an extensive database of actual bond transactions. The new model-based clustering approach was found to have tangible benefits both in providing portfolio diversification and intuitive insight into portfolio allocation decisions.
590 $a School code: 0047.
650 4 $a Applied Mathematics. $3 1669109
650 4 $a Engineering. $3 586835
690 $a 0364
690 $a 0537
710 2 $a The Claremont Graduate University. $b School of Mathematical Sciences. $3 2096458
773 0 $t Dissertations Abstracts International $g 79-12B.
790 $a 0047
791 $a Ph.D.
792 $a 2018
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10789848