However, to determine the clustering tendency, or
clusterability, is a hard task.
Just because there are so many definitions of clusters.
For example, we studied partitioning methods, hierarchical methods,
density-based methods, graph-based methods.
All these different methods may have different definitions of clusters.
That's why to study whether the data is clusterable or not is pretty hard.
However, even we just fix the cluster type.
For example, we just study the partitioning methods,
it is still hard to define an appropriate null model for a data set.
However, there are still some studies on clusterability assessment.
For example, there are methods like spatial histogram method,
distant distribution method, a Hopkins statistic.
The general philosophies, they try to compare their measure with the measure
generate from random samples, to see whether they are rather different.
For example, for spatial histogram method is try to contrast
the histogram of data with the histogram
generated from the random sample to see whether they are rather different.