# iterate through all words in this document split by space and create the vector of words
for word in documents[i].strip().split(' '):
if len(word) > 0: # if this is a word (i.e. not a null character), we add it to the vector
vector[ dictionary[word] ] = 1 # the word is in this document, so we set it to 1. For the
# index, we pull it from the word dictionary we created above
# update the current document we're processing with the new vector of words contained in that document
documents[i] = vector
# Create Gaussian NB classifier and fit it to the data
gnb = GaussianNB() # create the classifier object
gnbfit = gnb.fit(documents, labels) # fit (/train) the classifier to the data
pred = gnbfit.predict(documents) # classify the documents again using the classifier to test
# for accuracy
print len(dictionary),"words in the word dictionary"
print len(documents),"data points in the training set"
print "f1-score (weighted) of classifier on training set is:", f1_score(labels, pred, average='weighted')
f = open('test.txt', 'r')
test = cleanBody.cleanBody(f.readline())
vector = [0] * len(dictionary)
for word in test.strip().split(' '):
if len(word) > 0 and word in dictionary:
vector[ dictionary[word] ] = 1
prediction = gnbfit.predict(vector)
print "Predicted as category: %s (id:%d)" % (categories[prediction], prediction)
# basketball (3)
# nil (4)
max_per_category = 100
c = 0
print "Size of articles:", len(articles)
# for each article in articles
for arturl in articles:
# if all categories have more than max_per_category then
# quit looping. nil category excluded
print count[2:4]
if min(count[2:4]) < max_per_category:
c += 1
print c, "..", arturl
category,content = getPage.getPage(arturl)
if category == False and content == False: # no content on this page? add it to the nil category
count[4] += 1
else:
if count[category - 1] < max_per_category:
count[category - 1] += 1
data.append( "%d %s" % (category,cleanBody.cleanBody(content)) )
else:
break
print data
print count
file = open("data.txt", "w")
file.write("\n".join(data).encode('utf-8'))
word]] = 1 # the word is in this document, so we set it to 1. For the
# index, we pull it from the word dictionary we created above
# update the current document we're processing with the new vector of words contained in that document
documents[i] = vector
# Create Gaussian NB classifier and fit it to the data
gnb = GaussianNB() # create the classifier object
gnbfit = gnb.fit(documents, labels) # fit (/train) the classifier to the data
pred = gnbfit.predict(
documents) # classify the documents again using the classifier to test
# for accuracy
print len(dictionary), "words in the word dictionary"
print len(documents), "data points in the training set"
print "f1-score (weighted) of classifier on training set is:", f1_score(
labels, pred, average='weighted')
f = open('test.txt', 'r')
test = cleanBody.cleanBody(f.readline())
vector = [0] * len(dictionary)
for word in test.strip().split(' '):
if len(word) > 0 and word in dictionary:
vector[dictionary[word]] = 1
prediction = gnbfit.predict(vector)
print "Predicted as category: %s (id:%d)" % (categories[prediction],
prediction)
