I'm trying to analyse a dataset with the k-means algorithm in python with scikit-learn. Has anyone ever had the problem of getting less clusters than requested? With the input from sklearn.cluster import KMeans kmeans = KMeans(n_clusters=4) my samples are only in cluster 0, 1 and 2. Cluster 3 never appears. Doesn't the algorithm recognise that it could take any sample as new cluster(-center) when there is an empty cluster in the end? Continue reading...
