def test_logistic_regression(X , y , test_size = 0.3 , iters = 10 , num_folds = 5):
# if os.path.isfile('models/lr.model.p'):
# return pickle.load(open('models/lr.model.p'))
lr = LogisticRegression()
grid_params = { 'C':[0.01,0.1,1.0,10,100]}
clf = perform_classification(lr , grid_params , X , y , test_size , iters, num_folds )
# pickle.dump(clf , open('models/lr.model.p','wb'))
return clf
def test_support_vectors(X , y , test_size = 0.3 , iters = 10 , num_folds = 4):
# if os.path.isfile('models/svm.model.p'):
# return pickle.load(open('models/svm.model.p'))
svm = SVC()
# svm = LinearSVC()
# grid_params = {'C': [1, 10, 100, 1000], 'kernel': ['linear','rbf']}
# {'C': [1, 10, 100, 1000], 'gamma': [0.001, 0.0001], 'kernel': ['rbf']}]
grid_params = {'C': [1, 10, 100, 1000]}
clf = perform_classification(svm , grid_params , X , y , test_size , iters, num_folds )
# pickle.dump(clf , open('models/svm.model.p','wb'))
return clf
def test_random_forests(X , y , test_size = 0.3 , iters = 20 , num_folds = 5):
# if os.path.isfile('models/rf.model.p'):
# return pickle.load(open('models/rf.model.p'))
# X = np.delete(X , 8 , 1)
# print X.shape
rf = RandomForestClassifier()
grid_params = { 'n_estimators': [10,20,30],
'max_depth': [10,20,30],
'min_samples_split': [2 , 3 , 4 ]}
clf = perform_classification(rf , grid_params , X , y , test_size , iters, num_folds )
feature_order = [i[0]+1 for i in sorted(enumerate(clf.best_estimator_.feature_importances_), key=lambda x:x[1] , reverse=True)]
print feature_order
# pickle.dump(clf , open('models/rf.model.p','wb'))
return clf
