trend that are rather similar to EDS whereas C3 and C6 has although positive but slightly
different trend than EDS. The hourly trend for C4 is different from every other cluster
and EDS as it shows a negative trend. We could see that C4 also has a negative trend
for monthly analysis. Hence, our results show that using cluster analysis as a preliminary
task for accident data analysis can surely results in unknown findings which are
very difficult if only whole data set is analyzed. Also, use of cluster analysis as an initial
task for any accident data analysis removes the heterogeneity of the data to some extent,
which makes further analysis of the data easier. There our findings have the same opinion
with past studies [2, 3, 20, 22] that in order to improve the homogeneity in the data,
it is advisable to perform clustering on the road accident data set being used for analysis.