Exploring Region Classifier Based on Kolmogorov Complexity

Abstract

In this paper we explore the development of a parameter free region classifier based on Kolmogorov complexity. Given a set of regions described by unlimited but fixed number of attributes for each region, the region classifier will be able to build a classification tree which will help identify which regions are similar/dissimilar to each other based on a relative distance measure derived from Kolmogorov complexity. The region classifier is tested with the block level US Census demographics data as well as hitech establishment data for a subset of metropolitan regions. Preliminary results are presented for the census data as well as for the hitech sector for three different time periods.

Keywords: Kolmogorov Complexity, Regional classification, Hitech

JEL Classification: O18, R19