英文链接：http://scikit-/stable/auto_examples/applications/topics_extraction_with_nmf_lda.html
这是一个使用NMF和LDA对一个语料集进行话题抽取的例子。
输入分别是是tf-idf矩阵（NMF）和tf矩阵（LDA）。
输出是一系列的话题，每个话题由一系列的词组成。
默认的参数（n_samples/n_features/n_topics）会使这个例子运行数十秒。
你可以尝试修改问题的规模，但是要注意，NMF的时间复杂度是多项式级别的，LDA的时间复杂度与（n_samples*iterations）成正比。
几点注意事项:
（1）其中line 61的代码需要注释掉，才能看到输出结果。
（2）第一次运行代码，程序会从网上下载新闻数据，然后保存在一个缓存目录中，之后再运行代码，就不会重复下载了。
（3）关于NMF和LDA的参数设置，可以到sklearn的官网上查看【NMF官方文档】【LDA官方文档】。
（4）该代码对应的sk-learn版本为 scikit-learn 0.17.1
代码：
1 # Author: Olivier Grisel <olivier.grisel@> 2 # Lars Buitinck <L.J.Buitinck@uva.nl> 3 # Chyi-Kwei Yau <chyikwei.yau@> 4 # License: BSD 3 clause 5 6 from __future__ import print_function 7 from time import time 8 9 from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer10 from sklearn.decomposition import NMF, LatentDirichletAllocation11 from sklearn.datasets import fetch_20newsgroups12 13 n_samples = 200014 n_features = 100015 n_topics = 1016 n_top_words = 18 19 def print_top_words(model, feature_names, n_top_words):20for topic_idx, topic in enumerate(ponents_):21 print("Topic #%d:" % topic_idx)22 print(" ".join([feature_names[i]23for i in topic.argsort()[:-n_top_words - 1:-1]]))24print()25 26 27 # Load the 20 newsgroups dataset and vectorize it. We use a few heuristics28 # to filter out useless terms early on: the posts are stripped of headers,29 # footers and quoted replies, and common English words, words occurring in30 # only one document or in at least 95% of the documents are removed.31 32 print("Loading dataset...")33 t0 = time()34 dataset = fetch_20newsgroups(shuffle=True, random_state=1,35remove=('headers', 'footers', 'quotes'))36 data_samples = dataset.data37 print("done in %0.3fs." % (time() - t0))38 39 # Use tf-idf features for NMF.40 print("Extracting tf-idf features for NMF...")41 tfidf_vectorizer = TfidfVectorizer(max_df=0.95, min_df=2, #max_features=n_features,42stop_words='english')43 t0 = time()44 tfidf = tfidf_vectorizer.fit_transform(data_samples)45 print("done in %0.3fs." % (time() - t0))46 47 # Use tf (raw term count) features for LDA.48 print("Extracting tf features for LDA...")49 tf_vectorizer = CountVectorizer(max_df=0.95, min_df=2, max_features=n_features,50 stop_words='english')51 t0 = time()52 tf = tf_vectorizer.fit_transform(data_samples)53 print("done in %0.3fs." % (time() - t0))54 55 # Fit the NMF model56 print("Fitting the NMF model with tf-idf features,"57 "n_samples=%d and n_features=%d..."58 % (n_samples, n_features))59 t0 = time()60 nmf = NMF(n_components=n_topics, random_state=1, alpha=.1, l1_ratio=.5).fit(tfidf)61 exit()62 print("done in %0.3fs." % (time() - t0))63 64 print("\nTopics in NMF model:")65 tfidf_feature_names = tfidf_vectorizer.get_feature_names()66 print_top_words(nmf, tfidf_feature_names, n_top_words)67 68 print("Fitting LDA models with tf features, n_samples=%d and n_features=%d..."69 % (n_samples, n_features))70 lda = LatentDirichletAllocation(n_topics=n_topics, max_iter=5,71 learning_method='online', learning_offset=50.,72 random_state=0)73 t0 = time()74 lda.fit(tf)75 print("done in %0.3fs." % (time() - t0))76 77 print("\nTopics in LDA model:")78 tf_feature_names = tf_vectorizer.get_feature_names()79 print_top_words(lda, tf_feature_names, n_top_words)
结果：
Loading dataset...done in 2.222s.Extracting tf-idf features for NMF...done in 2.730s.Extracting tf features for LDA...done in 2.702s.Fitting the NMF model with tf-idf features,n_samples=2000 and n_features=1000...done in 1.904s.Topics in NMF model:Topic #0:don just people think like know good time right ve say did make really way want going new year llTopic #1:windows thanks file card does dos mail files know program use advance hi window help software looking ftp video pcTopic #2:drive scsi ide drives disk controller hard floppy bus hd cd boot mac cable card isa rom motherboard mb internalTopic #3:key chip encryption clipper keys escrow government algorithm security secure encrypted public nsa des enforcement law privacy bit use secretTopic #4:00 sale 50 shipping 20 10 price 15 new 25 30 dos offer condition 40 cover asking 75 01 interestedTopic #5:armenian armenians turkish genocide armenia turks turkey soviet people muslim azerbaijan russian greek argic government serdar kurds population ottoman millionTopic #6:god jesus bible christ faith believe christians christian heaven sin life hell church truth lord does say belief people existenceTopic #7:mouse driver keyboard serial com1 port bus com3 irq button com sys microsoft ball problem modem adb drivers card com2Topic #8:space nasa shuttle launch station sci gov orbit moon earth lunar satellite program mission center cost research data solar marsTopic #9:msg food chinese flavor eat glutamate restaurant foods reaction taste restaurants salt effects carl brain people ingredients natural causes olneyFitting LDA models with tf features, n_samples=2000 and n_features=1000...done in 22.548s.Topics in LDA model:Topic #0:government people mr law gun state president states public use right rights national new control american security encryption health unitedTopic #1:drive card disk bit scsi use mac memory thanks pc does video hard speed apple problem used data monitor softwareTopic #2:said people armenian armenians turkish did saw went came women killed children turkey told dead didn left started greek warTopic #3:year good just time game car team years like think don got new play games ago did season better llTopic #4:10 00 15 25 12 11 20 14 17 16 db 13 18 24 30 19 27 50 21 40Topic #5:windows window program version file dos use files available display server using application set edu motif package code ms softwareTopic #6:edu file space com information mail data send available program ftp email entry info list output nasa address anonymous internetTopic #7:ax max b8f g9v a86 pl 145 1d9 0t 34u 1t 3t giz bhj wm 2di 75u 2tm bxn 7eyTopic #8:god people jesus believe does say think israel christian true life jews did bible don just know world way churchTopic #9:don know like just think ve want does use good people key time way make problem really work say need
