K均值预测和拟合预测

Question

我有一个简单的K-means程序，该程序提取2个聚类，然后尝试预测新句子。我想为每个群集找到最佳的“适合”。

在我的“示例预测”中，_c0非常适合聚类0，而“ predict_bad_fit”涵盖了聚类0和1

我想我必须为每个预测句子计算出与聚类质心的平均差异。我怎么做？。

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.cluster import KMeans
import numpy as np

c0_sents = ['cats and dogs','i like cats','cats not like dogs','cats and dogs animals',]
c1_sents = ['computer is for typing','i play games on my computer','programs run on computer','computer has screen']

vectorizer = TfidfVectorizer()
X = vectorizer.fit_transform(c0_sents+c1_sents)

k_means = KMeans(n_clusters=2)
k_means.fit(X)

order_centroids = k_means.cluster_centers_.argsort()[:, ::-1]
terms = vectorizer.get_feature_names()
for i in range(2):
    for ind in order_centroids[i, :4]:
        print (i, terms[ind])

#0 computer
#0 on
#0 screen
#0 has
#1 cats
#1 dogs
#1 like
#1 and

predict_c0 = ['cats are not dogs']
predict_c1 = ['typing on computers']
predict_bad_fit = ['cats on computers dogs on screen']

for sent in predict_c0+predict_c1+predict_bad_fit:
    X = vectorizer.transform([sent])
    predicted = k_means.predict(X)
    print (sent,predicted)

#cats are not dogs [0]
#typing on computers [1]
#cats on computers dogs on screen [1]