robust single linkage failed to combine clusters that closer than the threshold
Z-Qing opened this issue · 0 comments
Z-Qing commented
Could anyone please tell me why the alt_labels generated by following codes will have clusters that closer than 1?
clusterer = hdbscan.RobustSingleLinkage(cut, k)
cluster_labels = clusterer.fit_predict(data)
hierarchy = clusterer.cluster_hierarchy_
alt_labels = hierarchy.get_clusters(1, 4)
data is coordinates of about 10 million 2d points. I already changed the parameter cut from 0.2 to 2 and k from 7 to 100, which didn't help. Also clusters in my data don't have any structures. Their shapes are just circles and ellipses.