Hierarchical agglomerative cluster analysis with a contiguity constraint
Brant H Wipperman · Summit (Simon Fraser University) · 2004
Cluster analysis is a technique for finding group structure in data; it is a branch of multivariate statistics which has been applied in many disciplines.The most common method of cluster analysis is hierarchical agglomeration.Several algorithms are discussed, with a focus on complete linkage.Constrained classification is then presented, specifically the case in which members of a cluster are required to be geographically contiguous.An example is provided, illustrating the creation of territories for automobile insurance in British Columbia, Canada.The dissimilarities between objects are measured by symmetrized deviance drops.This approach may be described as model-based clustering subject to contiguity constraints.First of all, I would like to thank my supervisors.It seems fitting to acknowledge Michael Stephens in a list format.I am grateful for his following contributions: (a) friendship; (b) house-sitting opportunities; (c) generous research assistantships; (d) embracing my project idea; (e) enduring numerous Microsoft Word glitches; (f) tolerating my German writing style, with its frequently recurring adjectives; (g) patience with my work schedule.Richard Lockhart has been personally responsible for more than half of my Masters credits.He was extremely understanding of my situation and