abstract
- © 2019 The Authors. Published by Elsevier B.V.Process monitoring for quality is a big data-driven quality philosophy aimed at defect detection through binary classification. The l1-regularized logistic regression learning algorithm has been successfully applied in manufacturing systems for rare quality event detection. Since the optimal value of the regularization parameter is not known in advance, many models need to be created and tested to find the final model to be deployed at the plant. In this context, model selection becomes a critical step in the process of developing a manufacturing functional model. Since most mature organizations generate only a few defects per million of opportunities, a novel model selection criterion (3D - LR) is introduced aimed at analyzing highly/ultra unbalanced binary data structures. Proposed criterion combines three of the most important attributes - prediction, fit, complexity - of each candidate model and maps them into a 3D space to select the best one. According to empirical results, proposed criterion outperforms - in parsimony and detection - Akaike information criterion, Bayesian information criterion, validation cross-entropy error and McFadden's adjusted R-squared, four of the most widely used criteria.