Χ2- clustering method in the problems of marketing

Abstract

The need for data clustering occurs in many practical problems,in particular in the problems of marketing. The similar objects grouping allows to study each group with a unified position, for example, we can propose a unified marketing strategy. There are many clustering methods. Each of these methods can produce the results that are different from others. Thus, it is believed that the clustering method is the only preliminary, exploratory data analysis. We propose a method of clustering on the basis of a comparison of empirical and theoretical distributions using χ2 consistency test. This test allows, regardless of the population distribution, statistically valid grouping of multidimensional data. That is, we can realize clustering with some prescribed probability of error. The method is implemented in Delphi software environment