Multicriterion Clusterwise Regression for Joint Segmentation Settings: An Application to Customer Value

Michael J. Brusco, Florida State University
Dennis J. Cradit, Florida State University
Armen Tashchian, Kennesaw State University


The authors present a multicriterion clusterwise linear regression model that can be applied to a joint segmentation setting. The model enables the consideration of segment homogeneity, as well as multiple dependent variables (segmentation bases), in a weighted objective function. The authors propose a heuristic solution strategy based on simulated annealing and examine trade-offs in the recovery of multiple true cluster structures for several synthetic data sets. They also propose an application of the model to a joint segmentation problem in the telecommunications industry, which addresses important issues pertaining to the selection of the objective function weights and the number of clusters.