Research Information System
3rd International Conference on Social and Educational Sciences, İstanbul, Turkey, 19 - 21 October 2023, pp.30-31
There are two main clustering methods, including
hierarchical and non-hierarchical methods. In cases where the number of
clusters is not known, hierarchical clustering method is used. In hierarchical
clustering, the number of clusters are discovered by using a graphical tool
called dendogram. The number of clusters that is represented with the letter
“k” is determined from dendogram by cutting across the branches at a given
level of distance measured by one of the axes. In non-hierarchical methods, the
number of clusters is predetermined. General approach for the predetermination
of “k” is running the hierarchical clustering method to determine the “k” value
at first and then using this value in the non-hierarchical methods. Next, it is
checked whether the cluster members obtained from these two methods are
similar. However, it is not possible to
automatically determine the number of clusters using either of these two
methods. On the other hand, using a third method, called the two-stage
clustering method, the number of clusters can be determined automatically and
the findings can be compared with the previous two methods. This study is a
study in which the mentioned three cluster methods were utilized in order to
find out the exact cluster solution. By doing so, the analyses enabled us to
approve the existence of the exact number of clusters. In the present study,
traffic accident indicators (total number of accidents, total number of road
motor vehicles, total number of persons death and number of persons injured) in
Turkey in 2022 by province-based were analyzed through clustering analysis. The
data are taken from the official website of the Turkish Statistical Institute
and analyzed with the SPSS program. All three mentioned methods were performed
with the traffic accident data and appropriate number for clusters were
determined. All three methods concludes that a two-cluster solution is the best
solution for the traffic accident data. The province of Istanbul, which is in
the first cluster, differs from the rest of the provinces in the second
cluster. According to traffic accident indicators, it has been determined that
the province of Istanbul is significantly separated from all other provinces.