DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Kim, Sung-Ho | - |
dc.contributor.advisor | 김성호 | - |
dc.contributor.author | Lee, Joo-Min | - |
dc.contributor.author | 이주민 | - |
dc.date.accessioned | 2011-12-14T04:54:11Z | - |
dc.date.available | 2011-12-14T04:54:11Z | - |
dc.date.issued | 2002 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=173586&flag=dissertation | - |
dc.identifier.uri | http://hdl.handle.net/10203/42047 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 응용수학전공, 2002.2, [ [ii], 34 p. ] | - |
dc.description.abstract | Data mining (Berry and Linoff, 1997; Han and Kamber, 2001) is the process of uncovering previously unknown patterns and relationships in a large database using sophisticated statistical analysis and modeling techniques. The regression model, the decision tree and the neural network are the representative models for predictive modeling. These models have different characteristics and each has advantages and disadvantages. The regression model has several advantages including the ease of interpretation and the capability of representing the linear structure very well. But the regression model has disadvantages which are that this model assumes the linearity between input variables and the target variable and the independence of the input variables. The decision tree has several advantages including the ease of interpretation, the ability to model complex input/target associations and the ability automatically handle missing values without imputation. But the decision tree is less appropriate to predict the value of a continuous variable and the small perturbations in a train data set can sometimes have large effects on the structure of the tree. The neural network has several advantages including the versatility for approaching problems, the capability of producing good results in complicated domains and the capability of handling both continuous variables and categorical variables. But a drawback of the neural network is difficulty of interpretation of the model structure. We investigate important properties of each model through analyzing with real data. If any input variable containing missing values has a much effect on predicting a target variable, then the decision tree performs much better than the other two models. We examine the three models for credit scoring to illustrate this property. A drawback of the neural network is difficulty of interpretation. A reasonable effort for structure interpretation is using an approximation model for the model via the... | eng |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | neural network | - |
dc.subject | decision tree | - |
dc.subject | regression model | - |
dc.subject | data mining | - |
dc.subject | assessment of the models | - |
dc.subject | 모델평가 | - |
dc.subject | 신경망분석 | - |
dc.subject | 의사결정나무 | - |
dc.subject | 회귀분석 | - |
dc.subject | 데이터 마이닝 | - |
dc.title | (A) comparison of decision support tools | - |
dc.title.alternative | 의사결정 지원 도구들의 비교 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 173586/325007 | - |
dc.description.department | 한국과학기술원 : 응용수학전공, | - |
dc.identifier.uid | 020003423 | - |
dc.contributor.localauthor | Kim, Sung-Ho | - |
dc.contributor.localauthor | 김성호 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.