Abstract
Manufacturing process data collected over time are considered time-series data and can be arranged into control charts. Important applications can be centered around these data like, for example, recognition of specific patterns, pattern similarity, detecting anomalies, and clustering and classification of patterns. We study and evaluate a number of classification techniques for process control data. For pattern similarity, we examine distance measure with raw data and with new feature extracted from the data. The evaluation is conducted with common benchmark process control data for time series process variables. This paper shows that data mining and machine learning can be extremely beneficial in acquiring and producing knowledge and discoveries form process data to benefit the industry.