Data Mining based on Regression Diagnostics

• 발행기관: 한국자료분석학회
• 저자명: Kwangsik Oh
• 간행물 정보: 『Journal of The Korean Data Analysis Society (JKDAS)』Vol.6 No.6, 1479~1488쪽, 전체 10쪽
• 주제분류: 자연과학 > 통계학
• 파일형태: PDF
• 발행일자: 2004.12.30

영문 초록

Among the data mining techniques, model building for regression should be based on the regression diagnostics such as selection of influent variables or observations. But the existing regression method is not practical for the huge data set. In this paper, a new regression diagnostic method for detecting influential observations which is useful for huge data sets is proposed. The proposed method is compared to the existing diagnostic method through simulations.

목차

1. Introduction

2. Classical Regression Diagnostics

3. Proposed Regression Diagnostic

4. Numerical Study

5. Concluding Remarks

References

키워드

Influential observation Outlier detection Regression diagnostic Data mining Least squares support vector machine