I am trying to find clustering algorithms which can work with cosine similarity for tweet classification.
1
-
4$\begingroup$ Methods that operate directly with distances can work out of the box. This includes various forms of agglomerative hierarchical clustering, DBSCAN, spectral clustering, graph theoretic algorithms that work based on graph cuts, etc. Other methods like k-means, k-medoids, etc. can also work, if implemented correctly for cosine distance. $\endgroup$– user20160Commented Mar 6, 2017 at 11:39
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Spherical k-means is the classical example. But really, any clustering algorithm that can take an arbitrary distance measure should be applicable, including DBSCAN; E.g., SciKit-Learn's implementation lets you choose the distance metric for DBSCAN, where you can plug in cosine similarity [1].
[1] http://scikit-learn.org/stable/modules/generated/sklearn.cluster.DBSCAN.html
