## Browse

다양한 앙상블 모델을 이용한 개인신용평가 모델 개발
Credit Evaluation Model Using Various Ensemble Models

DC Field Value Language
dc.contributor.author정민규-
dc.date.accessioned2020-05-07T05:54:34Z-
dc.date.available2020-05-07T05:54:34Z-
dc.date.issued2020-
dc.identifier.other000000160950-
dc.identifier.urihttp://dcollection.snu.ac.kr/common/orgView/000000160950ko_KR
dc.description학위논문(석사)--서울대학교 대학원 :자연과학대학 통계학과,2020. 2. Myunghee Cho Paik.-
dc.description.abstractHome credit 은행의 데이터를 이용해 개인 신용평가 모델을 만들 어 X일 이상 연체할 사람을 예측해 보았다. 전처리 과정은 계층 적 구조로 이루어진 데이터를 가공하여 상위 데이터를 기준으로 하위 데이터의 통계량을 만드는 것에서 시작하였다. 각 대출자마 다 주어진 다수의 대출, 그리고 각 대출마다 주어진 월별 정보를 활용하였고, 다중 대체법을 이용해 결측치를 대체했다. 그리고 불 균형 데이터 문제를 해결하고자 5-fold 교차 검증(cross validation)을 할 때마다 학습 데이터에 오버샘플링을 했다. 모델은 logistic, boosting, DNN을 사용했고, 각 모델을 최적화한 뒤에 두 가지의 간단한 앙상블 모델을 만들었다. 이러한 과정을 통해 신용 등급을 만들어 등급별 차이를 비교해 보았고, 대출 간 기간 차이 가 클수록 신용등급이 높다는 점을 발견할 수 있었다-
dc.description.abstractIn this paper, we created a personal credit rating model to predi ct who would be overdue by more than X days. The process b egan by processing the hierarchical data to generate statistics o f the lower data based on the upper data. Multiple loans were given for each lender, and monthly information was given for e ach loan, and multiple imputation was used to replace missing values. And to solve the unbalanced data problem, we oversamp led the training data every 5-fold cross validation. We used logi stic, boosting, and DNN model, and after optimizing each model, we created two simple ensemble models. Through this process, we made credit ratings and compared the differences between g rades.-
dc.description.tableofcontents제 1 장 도입 ································································· 1
제 2 장 데이터 ····························································· 3
2.1 데이터 소개 ·························································· 3
2.2 데이터 전처리 ······················································· 4
2.3 결측치 ·································································· 5
2.4 불균형 데이터 ······················································ 8
제 3 장 학습 ································································ 10
3.1 모델 ···································································· 10
3.2 검증 및 학습 ························································ 11
제 4 장 결과 분석 및 논의 ····································· 13
참고문헌 ····································································· 16
Abstract ···································································· 17
-
dc.language.isokor-
dc.publisher서울대학교 대학원-
dc.subject.ddc519.5-
dc.title다양한 앙상블 모델을 이용한 개인신용평가 모델 개발-
dc.title.alternativeCredit Evaluation Model Using Various Ensemble Models-
dc.typeThesis-
dc.typeDissertation-
dc.contributor.department자연과학대학 통계학과-
dc.description.degreeMaster-
dc.date.awarded2020-02-
dc.identifier.uciI804:11032-000000160950-
dc.identifier.holdings000000000042▲000000000044▲000000160950▲-
Appears in Collections:
College of Natural Sciences (자연과학대학)Dept. of Statistics (통계학과)Theses (Master's Degree_통계학과)
Files in This Item: