def train_vectorized(feats, Y, model_path=None, grid=False):
# Vectorize labels
labels = [ labels_map[y] for y in Y ]
Y = np.array( labels )
# Vectorize feature dictionary
vec = DictVectorizer()
X = vec.fit_transform(feats)
norm_mat( X , axis=0 , copy=False)
# Grid Search
if grid:
print 'Performing Grid Search'
clf = do_grid_search(X, Y)
else:
#clf = LinearSVC(C=0.1, class_weight='auto')
#clf = LogisticRegression(C=0.1, class_weight='auto')
clf = SGDClassifier(penalty='elasticnet',alpha=0.001, l1_ratio=0.85, n_iter=1000,class_weight='auto')
clf.fit(X, Y)
# Save model
if model_path:
with open(model_path+'.dict' , 'wb') as f:
pickle.dump(vec, f)
with open(model_path+'.model', 'wb') as f:
pickle.dump(clf, f)
# return model
return vec, clf
def predict_vectorized(feats, clf, vec):
# Vectorize feature dictionary
# NOTE: do not fit() during predicting
#'''
vectorized = vec.transform(feats)
norm_mat( vectorized , axis=0 , copy=False )
confidences = clf.decision_function(vectorized)
labels = clf.predict(vectorized)
labels = [ reverse_labels_map[y] for y in labels ]
with open('stuff-a','wb') as f:
pickle.dump(confidences,f)
with open('stuff-b','wb') as f:
pickle.dump(labels,f)
#'''
'''
with open('stuff-a','rb') as f:
confidences = pickle.load(f)
with open('stuff-b','rb') as f:
labels = pickle.load(f)
'''
# Adjust edge cases of negative/neutrals
adjusted = []
for l,c in zip(labels,confidences):
if l == 'negative':
# Bias predictions toward neutral unless very confident
if False: #((c[1]-c[2]) < 1) and (c[2] > 0):
#if ((c[1]-c[2]) < 1) and (c[2] > 0):
adjusted.append('neutral')
else:
#print c, '\t', c[1] - c[2]
adjusted.append(l)
else:
adjusted.append(l)
labels = adjusted
#print '\n\n'
#print clf
#print '\n\n'
#for k,v in vars(clf).items():
# print k
# print '\t', v
# print
#print
#print '\n\n'
#print clf.decision_function(vectorized)
#print '\n\n'
return labels
