Examination of the most suitable number of clusters in K-Means clustering in the case of thematic cartography applications

With the clustering analysis used for multivariate mapping, objects in which two or more subjects have similar properties are found. It was aimed in this thesis study to perform two applications with K-Means method, which is a non-hierarchical method among clustering analysis methods, to test the suitability of clusters formed through number of clusters being entered by the user, which is by the nature of the method used, and to determine the most suitable number of cluster. The most suitable number of clusters was determined using the Davies-Bouldin Index value. One of the most important criteria in this process is the Davies-Bouldin Index value. Distinctiveness of clusters increases with decreasing index value. Considering the calculated index values, the number of clusters that possesses the value close to zero is the most suitable one. The results were shown with the applications performed. In the first application performed with the data including number of vehicles per km2 and the corresponding amount of fuel used (m3) in 2019, the lowest index value was calculated for 4 clusters. In the second application, the lowest index value was calculated for 6 clusters using the data including the average amount of water used (m3) per capita and the amount of water drawn annually (m3/person-year). It would be useful to calculate the Davies-Boulding Index value when conducting cluster analysis with the K-Means method.