Using multiple correspondence analysis in the analysis of simulation experiments: a study of dynamic scheduling strategies

Abstract

In many cases, simulation is used to compare several scheduling strategies that may be applied to a system (considered as possible scenarios), with regard to several performance criteria. The simulation study of dispatching rules to job shop systems is a classical example. Unfortunately, simulation runs provide various tables of data, which can be difficult to read and to interpret. We discuss the use of multiple correspondence analysis to address this problem. This multivariate statistical technique allows certain relations between variables to be pointed out. The graphical overview of the available information which is provided facilitates the results interpretation and the comparison of strategies. To illustrate this approach, multiple correspondence analysis is used to study the application of well known dispatching rules to a simplified flow shop.