품질향상을 위한 데이터 전처리 프로세스에 관한 연구

다국어 초록 (Multilingual Abstract)

Recently, with the development of improved data acquisition systems such as sensors and fast communication systems, various types of data that could not be collected before can be obtained. The results of analysis using the collected data can provide real-time information on specific trends or changes in quality characteristics and process variables, leading to reduced process monitoring and controlling time and quality cost. However, errors may appear during the data collection process due to such problems as noise or process environment. Utilization of these error-contained data can waste resources and render analysis results useless.
This study aims to provide guidelines for appropriate pre-processing of manufacturing data. The data pre-processing approach is in the order of data integration, missing value processing, outlier processing, feature engineering, variable transformation, and data imbalance processing. The proposed data pre-processing approach is compared to six other data pre-processing methods, and shows its advantage with respect to the F1 score. The approach proposed in this study is expected to be helpful to quality practitioners who want to extract insight by applying statistical process control and machine learning methods with manufacturing quality data.

목차 (Table of Contents)

• 제Ⅰ장. 서론 1
• 제Ⅱ장. 문헌 연구 3
• 2.1 데이터 통합 3
• 2.2 결측치 처리 4
• 2.3 이상치 처리 4
• 2.4 데이터 변환 5
• 2.5 피처엔지니어링 6
• 2.6 데이터 불균형 처리 6
• 제Ⅲ장. 데이터 전처리 기법 및 분석 모델 종류 9
• 3.1 결측치 처리 9
• 3.2 이상치 처리 11
• 3.3 데이터 변환 12
• 3.4 피처엔지니어링 13
• 3.5 데이터 불균형 처리 16
• 3.6 분석 알고리즘 18
• 3.7 성능 평가지표 21
• 제Ⅳ장. 데이터 전처리 프로세스 적용지침 23
• 제Ⅴ장. 사례연구 31
• 5.1 데이터 세트 31
• 5.2 분석 방법 33
• 5.3 분석 결과 35
• 제Ⅵ장. 결론 및 추후 연구 과제 40
• 참고문헌 41