Abstract
Inspection cost of finished products is one of major concerns in mobile phone manufacturing industry. Manufacturers need an effective method to reduce inspection items while maintaining the quality of examination. However, traditional feature selection methods have poor classification ability to identify defective goods in imbalanced/skewed data. In this work, we present a novel approach to tackle this issue. Implementation results show that our proposed method not only has better ability to detect defective products, but also can significantly reduce test items without losing overall classification accuracy. Copyright © 2006 IFAC.