Abstract:

Turkey is divided into 7 regions depending on the cities’ geographic locations. Since the geographic properties of the cities belonging the same region are the same, socio-economical properties like populations, migration rates, annual incomes per person are expected to be similar. Some cities may not possess the same socio-economic structure with the rest of the cities that are from the same region but are assigned to the region anyway just because of geographical proximity. This study aims to find the cities which are in a sense exceptional in their regions. In order to eliminate the effect of the geographical proximity of the cities, not exact locations of the cities but the estimate locations obtained from multi-dimensional scaling are used. At the first hand, a k-means clustering algorithm which only depends on the geographical locations of the cities are used to form 7 clusters. Then, a decision tree analysis is used to form the clusters using both coordinates of the cities and socio-economical properties. Clusters obtained by k-means and decision tree analysis are then compared by themselves and with the real regions of Turkey and discussed.

Keywords: