자료유형 | 학위논문 |
---|---|
서명/저자사항 | Contributions to Classification and Regression Trees. |
개인저자 | Li, Yuanzhi. |
단체저자명 | The University of Wisconsin - Madison. Statistics. |
발행사항 | [S.l.]: The University of Wisconsin - Madison., 2017. |
발행사항 | Ann Arbor: ProQuest Dissertations & Theses, 2017. |
형태사항 | 95 p. |
기본자료 저록 | Dissertations Abstracts International 81-04B. Dissertation Abstract International |
ISBN | 9781088308486 |
학위논문주기 | Thesis (Ph.D.)--The University of Wisconsin - Madison, 2017. |
일반주기 |
Source: Dissertations Abstracts International, Volume: 81-04, Section: B.
Advisor: Loh, Wei-Yin. |
이용제한사항 | This item must not be sold to any third party vendors.This item must not be added to any third party search indexes. |
요약 | This dissertation consists of two parts. In the first part, we introduce the algorithm Multinomial Logistic Regression Tree. Logistic regression tree recursively partitions the data and fits a logistic regression at each partition. It combines logistic regression and tree model. The tree structure of the model can handle the nonlinear features automatically and the node logistic regression model can provide prediction of response class probabilities. Previous logistic regression algorithms are designed for binary response data only. Here we extend the model to multinomial response data. Ouralgorithm also supports exhaustive search for numerical cut point selection.In the second part, we compare the prediction accuracy of nine popular regression algorithms on data sets with missing values. To handle the missing data, we consider two approaches, the default method of the algorithm and missing data imputation. At low missing rate, regression tree GUIDE and M5 achieve the best and at high missing rate, tree ensemble GUIDE Forest and Random Forest have the best performance. A new method of GUIDE imputation is the best imputation method for most of theregression algorithms in our experiment. |
일반주제명 | Statistics. |
언어 | 영어 |
바로가기 |
: 이 자료의 원문은 한국교육학술정보원에서 제공합니다. |