I have used the Markov Clustering Algorithm (MCL) to cluster tweets, based on their similarity. However, I got a too high number of clusters, and most of the clusters have only one tweet. Any suggestions to reduce the number of clusters?
$\begingroup$
$\endgroup$
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
The best way to reduce the number of clusters produced by the Markov Clustering Algorithm (MCA) will depend on the specifics of your data and how you want to cluster it. However, some general tips that may help include:
One way is to pre-process your data to remove noise and outliers. This can help to make your data more amenable to clustering and make it more likely that meaningful patterns will emerge.
Some methods that may be effective in reducing the number of clusters produced by MCA include downsampling the dataset, using a smaller value for the clustering parameter (such as the number of iterations), or using a threshold value to filter out small clusters.
Another approach is to play with the MCL parameters, such as the inflation rate, to see if that has an effect on the number of clusters produced. try increasing its value, this will lead to fewer, but larger clusters.
Use a different similarity measure.
you could try running MCL multiple times with different random seeds to see if that makes a difference.
$\begingroup$
$\endgroup$
2
Depending on the implementation you're using, you can adjust the granularity, which will influence how many clusters you will get.
-
$\begingroup$ Thank you for your answer, you mean I adjust inflation parameter of mcl to get better clusers? Is that what you mean by granularity? $\endgroup$ Commented Sep 24, 2021 at 23:07
-
$\begingroup$ No, granularity is not the same as inflation. Are there no other parameter you can adjust? $\endgroup$ Commented Sep 25, 2021 at 18:26