def train(spamDir, hamDir):
featureset = []
for text in getDirContent(spamDir):
featureset.append((featuresForText(text), SPAM))
for text in getDirContent(hamDir):
featureset.append((featuresForText(text), NOSPAM))
shuffle(featureset)
classifier = NaiveBayesClassifier.train(featureset)
saveClassifier(classifier, CLASSIFIER_PATH)
print "Done with learning."
def getBayesAccuracy(splitRatio=0.9):
featureset = []
for text in getSpamContent():
featureset.append((featuresForText(text), SPAM))
for text in getHamContent():
featureset.append((featuresForText(text), NOSPAM))
shuffle(featureset)
trainset, devset = splitByRatio(featureset, splitRatio)
classifier = NaiveBayesClassifier.train(trainset)
print classifier.show_most_informative_features(10)
return nltk_classify.accuracy(classifier, devset)
def classify(evalDir, resultFilename):
classifier = loadClassifier(CLASSIFIER_PATH)
if not classifier:
raise Exception("Classifier was not loaded.")
print "loaded"
with open(resultFilename, "w") as f:
for text, filepath in iterDirContent(evalDir, yieldFilepath=True):
classification = classifier.classify(featuresForText(text))
f.write("%s\t%s\n" % (
filepath,
S_SPAM if classification else S_NOSPAM
))
print "Classified output was saved to file '%s'." % resultFilename
print "Done with classifying."
