了解分層聚類輪廓分數中的標簽。 ValueError：標簽數為1。有效值為2到n_samples-1（包括1和2）

Question

我計划使用最大平均輪廓分數來確定多個不同對象集的切點和聚類數。 不幸的是，我遇到了ValueError: Number of labels is 1. Valid values are 2 to n_samples - 1 (inclusive)對於我循環訪問的某些集合， ValueError: Number of labels is 1. Valid values are 2 to n_samples - 1 (inclusive) 。 我的循環從2個簇開始（至少這是我的意圖），所以我不確定為什么會出現此錯誤。

我知道，有一個類似的問題在這里 ，但這個與K-均值優惠和我在看的層次聚類。 那里的解決方案包括將循環限制為K> = 2，我相信我已經做到了（我的猜測是該解決方案將告訴我我沒有正確限制此限制）。

我編寫的代碼對於許多對象集都可以正常工作，但是對於其中一些對象卻遇到了問題。 在這里，我包括了一個距離矩陣，該距離矩陣在我以3個簇開始時起作用，而另一個在7個簇（但不小於7個簇）下起作用。

# Imports for the reproducible example:

import numpy as np
from scipy.spatial.distance import squareform
from scipy.cluster.hierarchy import fcluster, dendrogram, linkage
from sklearn import metrics

# Two of the matrices in question:

D1=np.array([[0, 0.93, 0.61, 0.81, 0.76, 0.77, 0.95, 0.91, 0.66, 0.77, 0.20, 0.55, 0.95, 0.77, 0.95, 0.72, 0.71, 0.94, 0.89, 0.17, 0.86, 0.84, 0.87], 
[0.93, 0, 0.92, 0.95, 0.94, 0.75, 0.83, 0.77, 0.94, 0.90, 0.20, 0.85, 0.84, 0.81, 0.82, 0.93, 0.91, 0.86, 0.91, 0.33, 0.83, 0.88, 0.86], 
[0.61, 0.92, 0, 0.62, 0.86, 0.65, 0.86, 0.91, 0.80, 0.71, 0.20, 0.65, 0.84, 0.81, 0.86, 0.72, 0.86, 0.87, 0.89, 0.33, 0.85, 0.85, 0.89], 
[0.81, 0.95, 0.62, 0, 0.78, 0.25, 0.90, 0.84, 0.67, 0.17, 0.20, 0.57, 0.77, 0.50, 0.87, 0.70, 0.68, 0.90, 0.76, 0.40, 0.69, 0.77, 0.70], 
[0.76, 0.94, 0.86, 0.78, 0, 0.74, 0.95, 0.92, 0.78, 0.56, 0.20, 0.82, 0.93, 0.76, 0.96, 0.40, 0.80, 0.92, 0.92, 0.33, 0.86, 0.82, 0.90], 
[0.77, 0.75, 0.65, 0.25, 0.74, 0, 0.78, 0.76, 0.50, 0.44, 0, 0.69, 0.72, 0.20, 0.82, 0.79, 0.33, 0.63, 0.75, 0.20, 0.80, 0.33, 0.70], 
[0.95, 0.83, 0.86, 0.90, 0.95, 0.78, 0, 0.92, 0.89, 0.90, 0.40, 0.86, 0.89, 0.79, 0.47, 0.96, 0.85, 0.85, 0.86, 0.17, 0.77, 0.84, 0.89], 
[0.91, 0.77, 0.91, 0.84, 0.92, 0.76, 0.92, 0, 0.88, 0.88, 0.20, 0.80, 0.27, 0.72, 0.93, 0.91, 0.88, 0.79, 0.89, 0.33, 0.68, 0.90, 0.51], 
[0.66, 0.94, 0.80, 0.67, 0.78, 0.50, 0.89, 0.88, 0, 0.69, 0.20, 0.57, 0.91, 0.55, 0.95, 0.77, 0.49, 0.89, 0.88, 0.20, 0.72, 0.88, 0.81], 
[0.77, 0.90, 0.71, 0.17, 0.56, 0.44, 0.90, 0.88, 0.69, 0, 0.20, 0.74, 0.87, 0.62, 0.90, 0.53, 0.43, 0.86, 0.89, 0, 0.71, 0.74, 0.87], 
[0.20, 0.20, 0.20, 0.20, 0.20, 0, 0.40, 0.20, 0.20, 0.20, 0, 0, 0, 0.25, 0.40, 0.20, 0.20, 0.40, 0.20, 0, 0, 0.20, 0], 
[0.55, 0.85, 0.65, 0.57, 0.82, 0.69, 0.86, 0.80, 0.57, 0.74, 0, 0, 0.86, 0.76, 0.89, 0.75, 0.62, 0.84, 0.78, 0, 0.53, 0.82, 0.73], 
[0.95, 0.84, 0.84, 0.77, 0.93, 0.72, 0.89, 0.27, 0.91, 0.87, 0, 0.86, 0, 0.79, 0.90, 0.91, 0.82, 0.82, 0.87, 0, 0.79, 0.84, 0.48], 
[0.77, 0.81, 0.81, 0.50, 0.76, 0.20, 0.79, 0.72, 0.55, 0.62, 0.25, 0.76, 0.79, 0, 0.84, 0.79, 0.38, 0.81, 0.83, 0.20, 0.56, 0.39, 0.83], 
[0.95, 0.82, 0.86, 0.87, 0.96, 0.82, 0.47, 0.93, 0.95, 0.90, 0.40, 0.89, 0.90, 0.84, 0, 0.94, 0.94, 0.88, 0.72, 0.40, 0.76, 0.87, 0.95], 
[0.72, 0.93, 0.72, 0.70, 0.40, 0.79, 0.96, 0.91, 0.77, 0.53, 0.20, 0.75, 0.91, 0.79, 0.94, 0, 0.82, 0.96, 0.91, 0.17, 0.86, 0.89, 0.94], 
[0.71, 0.91, 0.86, 0.68, 0.80, 0.33, 0.85, 0.88, 0.49, 0.43, 0.20, 0.62, 0.82, 0.38, 0.94, 0.82, 0, 0.86, 0.78, 0.20, 0.75, 0.85, 0.81], 
[0.94, 0.86, 0.87, 0.90, 0.92, 0.63, 0.85, 0.79, 0.89, 0.86, 0.40, 0.84, 0.82, 0.81, 0.88, 0.96, 0.86, 0, 0.50, 0.40, 0.56, 0.66, 0.54], 
[0.89, 0.91, 0.89, 0.76, 0.92, 0.75, 0.86, 0.89, 0.88, 0.89, 0.20, 0.78, 0.87, 0.83, 0.72, 0.91, 0.78, 0.50, 0, 0.20, 0.52, 0.71, 0.77], 
[0.17, 0.33, 0.33, 0.40, 0.33, 0.20, 0.17, 0.33, 0.20, 0, 0, 0, 0, 0.20, 0.40, 0.17, 0.20, 0.40, 0.20, 0, 0, 0.33, 0.20], 
[0.86, 0.83, 0.85, 0.69, 0.86, 0.80, 0.77, 0.68, 0.72, 0.71, 0, 0.53, 0.79, 0.56, 0.76, 0.86, 0.75, 0.56, 0.52, 0, 0, 0.82, 0.75], 
[0.84, 0.88, 0.85, 0.77, 0.82, 0.33, 0.84, 0.90, 0.88, 0.74, 0.20, 0.82, 0.84, 0.39, 0.87, 0.89, 0.85, 0.66, 0.71, 0.33, 0.82, 0, 0.76], 
[0.87, 0.86, 0.89, 0.70, 0.90, 0.70, 0.89, 0.51, 0.81, 0.87, 0, 0.73, 0.48, 0.83, 0.95, 0.94, 0.81, 0.54, 0.77, 0.20, 0.75, 0.76, 0]])

D2=np.array([[0, 0.91, 0.71, 0.34, 0.63, 0.93, 0.95, 0.96, 0.84, 0.88, 0.90, 0.40, 0.91, 0.89, 0.76, 0.95, 0.90, 0.92, 0.73, 0.95, 0.88, 0.77, 0.85], 
[0.91, 0, 0.75, 0.85, 0.38, 0.91, 0.29, 0.79, 0.95, 0.92, 0.96, 0.20, 0.94, 0.94, 0.61, 0.96, 0.91, 0.95, 0.36, 0.97, 0.97, 0.85, 0.93], 
[0.71, 0.75, 0, 0.76, 0.63, 0.93, 0.91, 0.95, 0.92, 0.94, 0.94, 0.40, 0.96, 0.95, 0.13, 0.92, 0.93, 0.96, 0.64, 0.91, 0.95, 0.88, 0.95], 
[0.34, 0.85, 0.76, 0, 0.50, 0.89, 0.92, 0.92, 0.80, 0.87, 0.93, 0.40, 0.95, 0.97, 0.81, 0.98, 0.89, 0.95, 0.60, 0.94, 0.73, 0.88, 0.93], 
[0.63, 0.38, 0.63, 0.50, 0, 0.38, 0.33, 0.38, 0.50, 0.50, 0.57, 0.25, 0.43, 0.50, 0.63, 0.43, 0.50, 0.17, 0, 0.50, 0.29, 0.50, 0.43], 
[0.93, 0.91, 0.93, 0.89, 0.38, 0, 0.89, 0.79, 0.81, 0.73, 0.91, 0.20, 0.69, 0.81, 0.91, 0.92, 0.90, 0.88, 0.33, 0.88, 0.90, 0.82, 0.85], 
[0.95, 0.29, 0.91, 0.92, 0.33, 0.89, 0, 0.88, 0.92, 0.93, 0.93, 0.40, 0.90, 0.95, 0.88, 0.95, 0.81, 0.93, 0.25, 0.95, 0.94, 0.85, 0.86], 
[0.96, 0.79, 0.95, 0.92, 0.38, 0.79, 0.88, 0, 0.89, 0.81, 0.91, 0.40, 0.85, 0.92, 0.97, 0.90, 0.51, 0.86, 0.42, 0.95, 0.92, 0.59, 0.79], 
[0.84, 0.95, 0.92, 0.80, 0.50, 0.81, 0.92, 0.89, 0, 0.65, 0.93, 0.40, 0.75, 0.94, 0.96, 0.93, 0.90, 0.62, 0.64, 0.95, 0.70, 0.84, 0.89], 
[0.88, 0.92, 0.94, 0.87, 0.50, 0.73, 0.93, 0.81, 0.65, 0, 0.94, 0.20, 0.65, 0.90, 0.97, 0.96, 0.90, 0.78, 0.64, 0.95, 0.68, 0.85, 0.92], 
[0.90, 0.96, 0.94, 0.93, 0.57, 0.91, 0.93, 0.91, 0.93, 0.94, 0, 0.40, 0.94, 0.49, 0.93, 0.68, 0.88, 0.90, 0.67, 0.70, 0.92, 0.82, 0.74], 
[0.40, 0.20, 0.40, 0.40, 0.25, 0.20, 0.40, 0.40, 0.40, 0.20, 0.40, 0, 0.20, 0.40, 0.40, 0.40, 0.40, 0, 0.25, 0.40, 0.40, 0.25, 0.25], 
[0.91, 0.94, 0.96, 0.95, 0.43, 0.69, 0.90, 0.85, 0.75, 0.65, 0.94, 0.20, 0, 0.93, 0.95, 0.94, 0.88, 0.53, 0.56, 0.96, 0.76, 0.84, 0.90], 
[0.89, 0.94, 0.95, 0.97, 0.50, 0.81, 0.95, 0.92, 0.94, 0.90, 0.49, 0.40, 0.93, 0, 0.94, 0.54, 0.91, 0.94, 0.50, 0.67, 0.97, 0.92, 0.80], 
[0.76, 0.61, 0.13, 0.81, 0.63, 0.91, 0.88, 0.97, 0.96, 0.97, 0.93, 0.40, 0.95, 0.94, 0, 0.94, 0.93, 0.96, 0.64, 0.95, 0.94, 0.92, 0.91], 
[0.95, 0.96, 0.92, 0.98, 0.43, 0.92, 0.95, 0.90, 0.93, 0.96, 0.68, 0.40, 0.94, 0.54, 0.94, 0, 0.88, 0.95, 0.60, 0.26, 0.95, 0.93, 0.79], 
[0.90, 0.91, 0.93, 0.89, 0.50, 0.90, 0.81, 0.51, 0.90, 0.90, 0.88, 0.40, 0.88, 0.91, 0.93, 0.88, 0, 0.95, 0.75, 0.92, 0.96, 0.39, 0.65], 
[0.92, 0.95, 0.96, 0.95, 0.17, 0.88, 0.93, 0.86, 0.62, 0.78, 0.90, 0, 0.53, 0.94, 0.96, 0.95, 0.95, 0, 0.60, 0.92, 0.39, 0.93, 0.94], 
[0.73, 0.36, 0.64, 0.60, 0, 0.33, 0.25, 0.42, 0.64, 0.64, 0.67, 0.25, 0.56, 0.50, 0.64, 0.60, 0.75, 0.60, 0, 0.60, 0.64, 0.50, 0.56], 
[0.95, 0.97, 0.91, 0.94, 0.50, 0.88, 0.95, 0.95, 0.95, 0.95, 0.70, 0.40, 0.96, 0.67, 0.95, 0.26, 0.92, 0.92, 0.60, 0, 0.91, 0.90, 0.90], 
[0.88, 0.97, 0.95, 0.73, 0.29, 0.90, 0.94, 0.92, 0.70, 0.68, 0.92, 0.40, 0.76, 0.97, 0.94, 0.95, 0.96, 0.39, 0.64, 0.91, 0, 0.88, 0.91], 
[0.77, 0.85, 0.88, 0.88, 0.50, 0.82, 0.85, 0.59, 0.84, 0.85, 0.82, 0.25, 0.84, 0.92, 0.92, 0.93, 0.39, 0.93, 0.50, 0.90, 0.88, 0, 0.58], 
[0.85, 0.93, 0.95, 0.93, 0.43, 0.85, 0.86, 0.79, 0.89, 0.92, 0.74, 0.25, 0.90, 0.80, 0.91, 0.79, 0.65, 0.94, 0.56, 0.90, 0.91, 0.58, 0]])

# I create the linkage matrix from each (these are probably easier to look at than the distance matrices):
L1=linkage(squareform(D1))
L2=linkage(squareform(D2))

# The loop I use to determine the value of the silhouette for each cluster size:

silhouette1=[]
for i in range(3,L1.shape[0]):
    temp=fcluster(L1,i,criterion='maxclust')
    s=metrics.silhouette_score(D1,temp,metric='euclidean')
    silhouette1.append(s)

silhouette2=[]
for i in range(7,L2.shape[0]):
    temp=fcluster(L2,i,criterion='maxclust')
    s=metrics.silhouette_score(D2,temp,metric='euclidean')
    silhouette2.append(s)


# The above loops work fine, as I have restricted the above to the range it seems to want. I am not sure why this works, but the following do not work:

silhouette1b=[]
for i in range(2,L1.shape[0]):
    temp=fcluster(L1,i,criterion='maxclust')
    s=metrics.silhouette_score(D1,temp,metric='euclidean')
    silhouette1b.append(s)

silhouette2b=[]
for i in range(2,L2.shape[0]):
    temp=fcluster(L2,i,criterion='maxclust')
    s=metrics.silhouette_score(D2,temp,metric='euclidean')
    silhouette2b.append(s)

我的理解是，出於某種原因（知道該原因可以回答我的問題），輪廓分數不會僅針對兩個聚類進行計算。 但是，該錯誤似乎暗示可以，但我沒有通過將i設置為2來正確定義2個群集。

Answer 1

編輯：

使用以下功能進一步診斷：

for i in range(1,len(D1)+1):
    print(str(i)+' clusters')
    r=fcluster(L1,i,criterion='maxclust')
    print(r)

顯示fcluster無法識別預期的群集數量。 看起來fcluster返回的簇可能少於“ maxclust”標准，這是有道理的。 我不確定我是否完全理解其原因，但是我希望這與某些元素的距離相同有關。

對於fcluster為什么會返回較少的群集的更多見解，將不勝感激。