Unsupervised adaptive resonance theory neural networks for control chart pattern recognition

Abstract

This paper describes the use of unsupervised adaptive resonance theory ART2 neural networks for recognizing patterns in statistical process control charts. To improve the classification accuracy, three schemes are proposed. The first scheme involves using information on changes between consecutive points in a pattern. The second scheme modifies the ART2 vigilance parameter during training. The third scheme merges class neurons representing the same class after training. The paper gives results which demonstrate the improvements achieved.