'k-means++' : selects initial cluster centroids using sampling based on an ... last iteration to make labels_ consistent with predict on the training set.
確定! 回上一頁