Capability of Control Chart Patterns Classifiers on Various Noise Levels

Abstract

Abstract Time Series Classification is one of the areas in data mining which receives some attention recently. Control Chart Patterns (CCPs) can be considered as time series. Monitoring and recognition of CCPs is also an importance process in manufacturing. This implies that ability to classify CCPs with high accuracy is essential. This study attempts to implement CCPs classifiers which are capable of dealing with CCPs with different level of noise. Extracting image processing statistical features is adopted as preprocessing technique. The work also investigates the effect of level of noise in classification. Three different types of techniques for implementing classifiers are selected, these are Decision Tree, Neural network and an evolutionary based program, known as Self-adjusting Association Rules Generator (SARG). It was found that SARG yielded the best performance among them. To date, this study is an attempt to classify particular model of CCPs with highest level of noise.