Cluster analysis of cross impact model scenarios

Analysis of outputs of cross impact models often focuses on counts of frequency of occurence of events, and frequency of co-occurence of events, to estimate the probabilities of these events. This paper describes an additional analysis which can be performed on the output of a cross impact model. Individual output scenarios are clustered together on the basis of similarity, utilizing standard techniques of cluster analysis. The resulting clusters of similar scenarios can be viewed as “typical” scenarios and analyzed in terms of how much they differ from the overall average for the complete set of output scenarios. Analysis of the reasons for these differences helps identify critical events and significant relationships among events. The paper discusses the clustering algorithm used and the type of output to be expected.