잠재합성변수를 활용하는 지도학습 모형

국문 초록 (Abstract)

군집분석은 주로 개체를 묶는 용도로 사용되지만 유사한 특성의 변수들을 군집화할 수도 있다. Chavent 외 3인 (2012)에 의해 개발된 R 패키지 ClustOfVar를 사용하면 변수의 유형에 제약없이 변수들을 군집으로 묶을 수 있다. 본 연구에서는 각
변수들의 군집에 PCAMIX 방법을 적용하여 구한 첫 번째 주성분 점수, 즉 잠재합성변수를 새로운 설명변수로 활용한 지도학습 모형을 제안한다. 또한 반응변수 예측력을 높이기 위해 PLS 회귀 방법론에서 착안한 잠재합성변수 보정 방법을 제안하고 이를 활용하여 지도학습을 수행한다.
본 연구에서 제안하는 PCA 수량화와 PLS 수량화 방법은 설명변수의 수가 클 때 특히 성능이 좋았다.

목차 (Table of Contents)

• 국문 초록 ··································································································· ⅰ
• 제 1 장 연구의 배경과 목적 ·································································· 1
• 제 2 장 제안 방법론 ··············································································· 2
• 2.1 PCA 수량화 ····························································································· 2
• 2.2 PLS 수량화 ····························································································· 6
• 제 3 장 사례 분석 ················································································· 8
• 3.1 분석 절차 ································································································· 8
• 3.2 적용 알고리즘 ························································································· 9
• 3.3 적용 사례 및 결과 ··············································································· 10
• 3.3.1 BreastCancer 자료 ········································································· 11
• 3.3.2 yarn 자료 ···························································································· 13
• 3.3.3 gasoline 자료 ····················································································· 16
• 3.3.4 leukemia 자료 ···················································································· 18
• 제 4 장 맺음말 ······················································································· 19
• 참고문헌 ····································································································· 20