K-Means 聚类中距质心最近的 M 个点

Question

我实现了一个函数，用于查找运行 K 均值聚类算法后计算出的每个质心的“最近”数据点。我想知道是否有一个 sklearn 函数可以让我找到距离每个质心最近的

M

点。

Answer 1

sklearn.neighbors.NearestNeighbors

来拟合我们的数据集。然后，我们可以使用 K 均值质心查询最近邻模型来检索邻居。像这样：

# Copyright 2024 Google LLC.
# SPDX-License-Identifier: Apache-2.0
from sklearn.cluster import KMeans
from sklearn.neighbors import NearestNeighbors

# random dense embeddings for 100 points with 10 dimensions.
dataset = np.random.rand(100,10)

# fit K-means with 3 clusters on our dataset.
kme = KMeans(n_clusters=3)
kme.fit(dataset)

# we should have 3 vectors for 3 centroids.
print(kme.cluster_centers_.shape) # (3, 10)

# initialize NearestNeighbor with 5 neighbors and fit our dataset.
knn = NearestNeighbors(n_neighbors=5, metric='cosine')
knn.fit(dataset)

# Use the model to query the centroids' neighbors.
distances, indices = knn.kneighbors(kme.cluster_centers_)

for centroid, distance_from_centroid, index in zip(kme.cluster_centers_, distances, indices):
  print(centroid, distance_from_centroid, index)

最后一个循环将输出 3 行。每个都类似于质心的向量以及其最近邻居的 5 个距离和索引。

Answer 2

http://scikit-learn.org/stable/modules/neighbors.html

类sklearn.neighbors.NearestNeighbors为您找到它：

http://scikit-learn.org/stable/modules/ generated/sklearn.neighbors.NearestNeighbors.html#sklearn.neighbors.NearestNeighbors