Statistical process control (SPC) techniques that originated in manufacturing have also been applied to monitoring the quality of various service processes, which can be characterized by one or several variables. Conventional multivariate SPC methods usually have an underlying distribution, generally multivariate normal, assumed for the process variables. With distribution parameters estimated, the whole structure of the data can then be determined. With just a few parameters, the distribution assumption offers information about the spreading condition of the data and thus reduces effort in describing it. However, in many cases, the distribution assumption cannot be easily made, or the assumption made is not appropriate. For instance, the quality characteristics of a service process may...[ Read more ]

Statistical process control (SPC) techniques that originated in manufacturing have also been applied to monitoring the quality of various service processes, which can be characterized by one or several variables. Conventional multivariate SPC methods usually have an underlying distribution, generally multivariate normal, assumed for the process variables. With distribution parameters estimated, the whole structure of the data can then be determined. With just a few parameters, the distribution assumption offers information about the spreading condition of the data and thus reduces effort in describing it. However, in many cases, the distribution assumption cannot be easily made, or the assumption made is not appropriate. For instance, the quality characteristics of a service process may include both continuous and categorical variables (i.e., mixed-type variables). In this case there will be no specific distribution to assume. Direct application of conventional SPC techniques to monitor such mixed-type variables may cause increased false alarm rates and misleading conclusions. To further complicate the case, the number of variables is usually large (i.e., high-dimensional variables).

This is indeed a common situation in reality. Three examples from service industries are given to demonstrate the universality of the mixed-type variables in a service process. This dissertation presents approaches for monitoring a process that is characterized by mixed-type variables. An improved design of the support-vector-machine-based k-charts is presented. Besides the distance-based method, a density-based scheme is also shown to be useful. Then, a LASSO-based approach is proposed when relationships exist among the mixed-type variables. Illustrations and comparisons are presented based on a real example from a logistics firm. The results confirm the improved performances from our proposed design schemes.