x_train=train_df.data[0:7000]
y_train=train_df.class_num[0:7000]

def my_tokenize(s):
    sbs =SnowballStemmer('english')
    return [sbs.stem(w) for  w in wordpunct_tokenize(s)]


vect = CountVectorizer(tokenizer=my_tokenize,stop_words='english',max_features=5000)
# learn training data vocabulary, then use it to create a document-term matrix
train_dtm =vect.fit_transform(x_train)
print (train_dtm)
#transform testing data (using fitted vocabulary) into a document-term matrix
test_dtm=vect.transform(x_test)
#instantiate a Multinomial Naive Bayes model
nb = MultinomialNB(fit_prior=True)
#train the model train_dtm
nb.fit(train_dtm, y_train)
# make class predictions for test_dtm
y_test = nb.predict(test_dtm)

run1
....
....
(6998, 2153) 1
(6998, 3908) 1
(6998, 183) 1
(6998, 1369) 3
(6998, 43) 1
(6998, 185) 2
(6998, 2389) 1
(6998, 2137) 1
(6998, 2757) 1
(6998, 555) 1
(6998, 847) 1
(6998, 853) 1
(6998, 2994) 1
(6998, 3012) 1
(6999, 3185) 1
(6999, 2292) 1
(6999, 402) 1
(6999, 43) 1
(6999, 4774) 1
(6999, 4566) 1
(6999, 2940) 1
(6999, 555) 1
(6999, 847) 1
(6999, 1562) 1
(6999, 1294) 1

run2
....
....
      (6997, 3889)  2
      (6997, 2971)  1
      (6997, 85)    2
      (6997, 55)    3
      (6997, 139)   9
      (6997, 3006)  2
      (6998, 2981)  1
      (6998, 3172)  1
      (6998, 43)    1
      (6998, 1) 1
      (6998, 2338)  1
      (6998, 4063)  1
      (6998, 3921)  1
      (6998, 545)   1
      (6998, 842)   1
      (6998, 2833)  1
      (6998, 3889)  1
      (6998, 139)   1
      (6999, 43)    1
      (6999, 1)     1
      (6999, 153)   1
      (6999, 545)   1
      (6999, 842)   1
      (6999, 4760)  1
      (6999, 3889)  1

run3
....
....
(6994, 2977) 1
(6994, 817) 2
(6994, 132) 1
(6994, 144) 1
(6995, 3946) 1
(6996, 2838) 1
(6996, 4858) 1
(6996, 2427) 1
(6996, 2153) 1
(6996, 3010) 1
(6997, 4995) 1
(6997, 4809) 1
(6997, 1824) 1
(6997, 4833) 1
(6998, 1307) 1
(6998, 4889) 1
(6998, 144) 2
(6999, 44) 1
(6999, 162) 1
(6999, 3428) 1
(6999, 551) 1
(6999, 836) 1
(6999, 2928) 1
(6999, 3288) 1
(6999, 3909) 1