用python代码实现1. 数据预处理：利用Jieba分词对每篇文档进行分词；利用TF-ID对F每篇文档提取关键特征词；利用特征词的TF-IDF值作为文档的特征向量。 2. 对预处理后的文档（每篇文档用特征向量表示），从每个类别中随机选取40篇文档，共得到40×5=200篇文档。 3. 对提取的200篇文档，采用K-means算法，划分5个簇，每个簇代表一个类别 4. 画出聚类结果散点图，同时计算FMI指数

以下是实现上述流程的Python代码：

数据预处理：

import jieba
from sklearn.feature_extraction.text import TfidfVectorizer

# 分词
def cut_words(text):
    words = jieba.cut(text)
    return " ".join(words)

# TF-IDF特征提取
def get_tfidf(corpus):
    vectorizer = TfidfVectorizer()
    tfidf = vectorizer.fit_transform(corpus)
    return vectorizer.get_feature_names(), tfidf.toarray()

corpus = ['这是一篇文档', '这是另一篇文档']
corpus_cut = [cut_words(text) for text in corpus]
feature_names, tfidf = get_tfidf(corpus_cut)

从每个类别中随机选取40篇文档：

import random

# 从每个类别中随机选取num篇文档
def random_sample(docs, labels, num):
    sample_docs, sample_labels = [], []
    for label in set(labels):
        docs_label = [doc for i, doc in enumerate(docs) if labels[i] == label]
        sample_docs_label = random.sample(docs_label, num)
        sample_docs.extend(sample_docs_label)
        sample_labels.extend([label]*num)
    return sample_docs, sample_labels

docs = ['文档1', '文档2', '文档3', '文档4', '文档5', '文档6', '文档7', '文档8', '文档9', '文档10']
labels = [0, 0, 0, 1, 1, 1, 2, 2, 2, 3]
sample_docs, sample_labels = random_sample(docs, labels, 2)

K-means聚类：

from sklearn.cluster import KMeans

# K-means聚类
def kmeans_clustering(features, n_clusters):
    kmeans = KMeans(n_clusters=n_clusters)
    kmeans.fit(features)
    return kmeans.labels_

features = [[1, 2], [3, 4], [5, 6], [7, 8], [9, 10]]
labels = kmeans_clustering(features, 2)

计算FMI指数：

from sklearn.metrics import fowlkes_mallows_score

# 计算FMI指数
def fmi_score(labels_true, labels_pred):
    return fowlkes_mallows_score(labels_true, labels_pred)

labels_true = [0, 1, 2, 0, 1]
labels_pred = [1, 0, 2, 1, 0]
fmi = fmi_score(labels_true, labels_pred)

画出聚类结果散点图：

import matplotlib.pyplot as plt

# 画出聚类结果散点图
def plot_clusters(features, labels):
    plt.scatter([f[0] for f in features], [f[1] for f in features], c=labels)
    plt.show()

features = [[1, 2], [3, 4], [5, 6], [7, 8], [9, 10]]
labels = [0, 1, 0, 1, 1]
plot_clusters(features, labels)