데이터파밍을 활용한 빅데이터 기반 품질개선 방법

Abstract

Machine learning has been utilized across various industries to optimize quality characteristics for quality improvement. However, existing methods such as Bayesian optimization and genetic algorithm suffer from drawbacks including theoretical complexity and a lack of knowledge about quality characteristics near the optimal conditions. This study proposes a data farming method to systematically find the optimal region of features by applying the Nearly Orthogonal Latin Hypercube(NOLH) design to quality big data. The proposed method employs easy-to-implement region-reduction technique by considering significant features for the labels. Moreover, unlike existing methods, it can present optimal features’ region ensuring a desired level of quality characteristics even if slight fluctuations occur in the process features. A case study shows that the proposed method performs better than other methods. The data farming method is expected to help practitioners to improve process performance using quality big data.