Market segmentation is a valuable analysis for businesses, especially when they have a diverse clientele. Clustering is the unsupervised learning method that can deal with market segmentation. While many clustering techniques exist, they generally suffer from instabilities and thus do not necessarily generate the best solutions to a problem. A novel method that does not suffer from these instabilities is convex clustering. Although convex clustering is already established for continuous and binary data, this paper extends the literature by generalising convex clustering for different data types. Two practical cases show that the generalised model has a practical use and can provide businesses with vital information. A simulation study furthermore reveals promising results: for categorical data this model outperforms the benchmark of an established method, especially when the relative difference between clusters is small. Finally, it is the first clustering method that is able to analyse Poisson distributed data and its performance in the simulation study is decent. Thus, this paper presents the generalisation for convex clustering for different data types and supports that this extension to the literature is logical and beneficial.

Additional Metadata
Thesis Advisor Koning, A.J.
Persistent URL
Series Econometrie
Kieskamp, T.T. (2020, February 14). Generalised Convex Clustering: A Parametric Technique to Perform Market Segmentation. Econometrie. Retrieved from