Abnormal Patterns Detection in Control Charts using Classification Techniques
Zalinda Othman, Huda Fathi Eshames - · International Journal of Advancements in Computing Technology · 2012
Any abnormal patterns show in Statistical Process Control charts imply the presence of possible assignable causes and variances that may lead to the process performance deterioration. Therefore, timely detection and recognizer of patterns in control charts are very important in the SPC implementation. This paper presents the performance of five classification methods on a set of large data for anomaly patterns detection in control charts. The control chart dataset has its specific features that need specific data preprocessing procedures. It is crucial and involves a number of stages of data preparation procedures. Firstly, the Principle Component Analysis is employed for similarity measure. Secondly, the Piecewise Aggregate Approximation and Symbolic Aggregate Approximation are used as data representation. The preprocessed data are fitted to the classification algorithms to extract important knowledge. The algorithms are support vector machine, decision tree, MLP networks, RIDOR algorithm and JRip algorithm. Numerical results showed that the JRip algorithm has the best performance compared to the others. It achieved highest detection accuracy about 99.66 % and the lowest error rate is 2.987.