Bayesian-ML/topic_modeling.py at master · markheimann/Bayesian-ML · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
#from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.datasets import fetch_20newsgroups
from sklearn.decomposition import NMF
import numpy as np
import lda
import time

#Can play around with all these parameters and see what happens.  Visualize topics in cool ways.
n_samples = 2000
n_features = 1000
n_topics = [2,4,6,8,10,20]
n_top_words = 10

#dataset: 20 newsgroups
#print "Loading dataset and extracting document term features..."
rand = np.random.mtrand.RandomState(8675309)
cats = ['rec.sport.baseball', 'sci.crypt']
dataset = fetch_20newsgroups(categories=cats,shuffle=True, random_state=rand, remove=('headers', 'footers', 'quotes'))
#dataset = fetch_20newsgroups(shuffle=True, random_state=rand, remove=('headers', 'footers', 'quotes'))

#get features from dataset
#vectorizer = TfidfVectorizer(max_df=0.95, min_df=2, max_features=n_features, stop_words='english')
#vectorizer = CountVectorizer(stop_words='english')
#td_matrix = vectorizer.fit_transform(dataset.data[:n_samples])
#feature_names = vectorizer.get_feature_names()
#vectorizer = CountVectorizer(stop_words='english')
vectorizer = CountVectorizer(min_df=10, stop_words='english')
td_matrix = vectorizer.fit_transform(dataset.data)
feature_names = vectorizer.get_feature_names()

#Perform NMF
#for reference: http://scikit-learn.org/stable/auto_examples/applications/topics_extraction_with_nmf.html
print("Performing NMF with %d samples and %d features..." % (n_samples, n_features))

for num in n_topics:
    # Time nmf
    start_time_NMF = time.time()
    nmf = NMF(n_components=num, random_state=rand).fit(td_matrix)
    print("NMF ran in %s seconds" % (time.time() - start_time_NMF))

    #print out topics
    print "NMF topics: "
    for topic_idx, topic in enumerate(nmf.components_):
        print("Topic #%d:" % topic_idx)
        print(" ".join([feature_names[i] for i in topic.argsort()[:-n_top_words - 1:-1]]))
        print

#Perform LDA
#For reference: https://pypi.python.org/pypi/lda

for num in n_topics:
    print("Performing LDA for %d topics..." % num)

    # Time LDA
    start_time_LDA = time.time()
    model = lda.LDA(n_topics=num, n_iter=200, random_state=rand)
    model.fit_transform(td_matrix)
    print("LDA ran in %s seconds" % (time.time() - start_time_LDA))

    #print out topics
    print "\nLDA topics: "
    for topic_idx, topic in enumerate(model.components_):
        print("Topic #%d:" % topic_idx)
        print(" ".join([feature_names[i] for i in topic.argsort()[:-n_top_words - 1:-1]]))
        print