# Using just CountVectorizer (threshold of 0.05 to 0.75) and Multinomial-Naive Bayes

import sys
from sklearn.naive_bayes import MultinomialNB
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.datasets import load_files
from sklearn.cross_validation import train_test_split
from sklearn import metrics
from sklearn.pipeline import Pipeline
import numpy

data_folder = "./data-hold/cleaned/"
dataset = load_files(data_folder, shuffle = False)

docs_train, docs_test, y_train, y_test = train_test_split(
    dataset.data, dataset.target, test_size=0.25, random_state=None)

pipeline = Pipeline([
    ('vect', CountVectorizer(min_df=0.05, max_df=0.75)),
    ('clf', MultinomialNB()),
])

pipeline.fit(docs_train, y_train)
y_predicted = pipeline.predict(docs_test)
print(metrics.classification_report(y_test, y_predicted, target_names = dataset.target_names))

################ Performance report
#                    precision    recall  f1-score   support

#    charles-m-blow       0.87      0.71      0.78        82
#      david-brooks       0.93      0.75      0.83       187
#       frank-bruni       0.69      0.83      0.76        60
#      gail-collins       0.84      0.89      0.87       169
#        joe-nocera       0.80      0.83      0.81        81
#      maureen-dowd       0.76      0.88      0.81       128
#  nicholas-kristof       0.84      0.83      0.83       123
#      paul-krugman       0.90      0.91      0.90       154
#       roger-cohen       0.80      0.86      0.83       115
#      ross-douthat       0.77      0.71      0.74        48
# thomas-l-friedman       0.87      0.87      0.87       116

#       avg / total       0.84      0.83      0.83      1263




## print out top 10 most informative features
clf = pipeline.steps[1][1]
vect = pipeline.steps[0][1]
for i, class_label in enumerate(dataset.target_names):
            topt = numpy.argsort(clf.coef_[i])[-10:]
            print("%s:    %s" % (class_label,
                  ", ".join(vect.get_feature_names()[j] for j in topt)))



# charles-m-blow:    republicans, said, some, our, most, those, were, obama, president, percent
# david-brooks:    new, government, now, them, over, some, these, do, were, obama
# frank-bruni:    last, many, just, re, them, had, him, said, her, she
# gail-collins:    get, state, mr, do, her, were, she, said, new, had
# joe-nocera:    do, them, years, other, said, she, new, were, its, had
# maureen-dowd:    hillary, even, were, him, had, president, said, her, obama, she
# nicholas-kristof:    my, also, some, because, our, year, said, had, her, she
# paul-krugman:    government, were, health, much, obama, economic, economy, now, even, mr
# roger-cohen:    had, states, united, american, now, israel, world, iran, obama, its
# ross-douthat:    well, because, new, many, even, just, party, our, its, obama
# thomas-l-friedman:    president, america, them, how, its, now, just, do, world, our