def multistage_classify(n, s, e, folds):
features, metafeatures, labels = get_data(n, s, e, metadata=True)
features2 = new_features(features, labels, metafeatures)
return kfold_crossval(npa(features2), npa(labels), folds, 2)
# v_features = (v.fit_transform(POS_features)).toarray()
# transform text into count vectors
v = CountVectorizer(min_df=1, max_features=2000)
v_features = (v.fit_transform(features)).toarray()
print 'Finished vectorizing text data'
X, Y = randomize(v_features, labels)
print len(X[0])
if len(X) == len(Y): print 'Data check ... OK'
else:
print 'Data check failed. Aborting execution'
return None
return kfold_crossval(X, Y, folds, 2, RF=True)
def iterateMNB(n_trials, d_range, s_size, folds):
C0L, C1L, WL = [], [], []
# upper and lower date limits
ll = datetime.date(2012,01,01)
ul = datetime.date(2013,06,01)
v = CountVectorizer(min_df=1, max_features=2000)
for i in range(n_trials):
s = tools.randomDate(ll, ul)
e = date.fromordinal(s.toordinal() + d_range)
# c0, c1, w = multistage_classify(s_size, s, e, folds)
c0, c1, w = classify_text(s_size, s, e, folds)