Abstract
1. Title: “Classification Trees to Interpret Out-Of-Control Signals in Multivariate Control Charts”
2. Objectives: In statistical quality control, one of the most widely used tools are the control charts. The main problem of the multivariate control charts, including the Hotelling‘s T2 control chart, lies in that they only indicate that a change in the process has happened, but they do not show which variable or variables are the source of this change. In the specialized literature there are many approaches to tackle this problem, although, the most usual consist on the decomposition of the T2 statistic. In this research, we propose an alternative method through the application of classification trees. This classifier is used to determine which variable or variables are the source of the change in the process.
3. Data/Methods: This paper uses simulated data to analyze whether classification trees are a good alternative to determine which variable or variables are the source of the change in the process.
4. Results/Expected results: The results have been compared with other approaches in the specialized literature and show that classification trees are a good alternative to interpret multivariate control charts.
.5. Conclusion: In this research, we propose an alternative method through the application of classification trees to interpret the multivariate control charts. This classification method is used to determine which variable or variables are the source of the change in the process. The results show that this method constitutes a good tool to help to interpret the multivariate control charts.
Key words: Statistical quality control, classification trees.