Preface
Traditional methods and techniques for identifying patterns in business data—such as K-means, self-organizing maps of Kohonen, and others—can achieve good results. However, they are usually focused on identifying individuals’ behaviors. They are based on a set of information about individuals and are clustered according to similar characteristics. All behaviors are placed in one of those cluster groups, so presumably individuals included in each group have similar characteristics and behaviors. However, as with any analytical model, it is an approximation of similar behaviors. Each group holds a particular average behavior. Each observation inside the group holds its own behavior. Some observations have behavior similar to the average group, some do not. By being aware that a clustering model is an approximation, any practical action should consider, for instance, the closest observations to the center of each cluster, assuring that these observations hold behavior similar to the cluster. Some particular clusters exhibit an average behavior. Some observations (customers) are close to this behavior, some are not. This is the approximation. The closer to the cluster’s center the observations are, the closer their individual behavior will be to the average cluster.
Social network analysis is also an approximation, but about the group rather than about the individual. The characteristics that are taken into consideration are not ones assigned to the individuals, but rather ...
