def train(self, splits, X=None, y=None):
"""train the CNN
Arguments:
splits {int array} -- split of data between types
Keyword Arguments:
X {ndarray} -- train input data (default: {None})
y {ndarray} -- train labels (default: {None})
"""
if not X or not y:
X = self.data['features']
y = self.data['labels']
X_train, y_train, X_val, y_val, X_test, y_test = split_data(
X, y, splits)
y_train = to_categorical(y_train)
y_val = to_categorical(y_val)
y_test = to_categorical(y_test)
self.model.fit(X_train,
y_train,
batch_size=128,
epochs=200,
verbose=2,
validation_data=(X_val, y_val))
self.test(X_test, y_test)
def train(self, splits, X=None, y=None, verbose=False):
"""train the svm
Arguments:
splits {float array} -- test train splits
Keyword Arguments:
X {ndarray} -- input data (default: {None})
y {ndarray} -- labels (default: {None})
verbose {bool} -- Print output (default: {False})
"""
if not X or not y:
X = self.data['features']
y = self.data['labels']
X_train, y_train, X_test, y_test = split_data(X, y, splits)
if verbose:
print("Feature Shape for SVM ", self.title)
print(X_train.shape)
print("Label Shape for SVM ", self.title)
print(y_train.shape)
self.model.fit(X_train, y_train)
if len(X_test) <= 0:
return
self.test(X_test, y_test, verbose)
def voter_metrics(self, splits, X=None, y=None, verbose=False):
'''
'''
if not X or not y:
X = self.data['features']
y = self.data['labels']
X_train, y_train, X_test, y_test = split_data(X, y, splits)
true_neg =0
false_neg =0
true_pos =0
false_pos =0
for i,X_i in enumerate(X_test):
pred = self.predict(X_i.reshape(1, -1))
true = int(y_test[i])
pred = int(pred)
if true == 0 and pred == 0:
true_neg += 1
elif true == 0 and pred == 1:
false_pos += 1
elif true == 1 and pred == 0:
false_neg += 1
elif true == 1 and pred == 1:
true_pos += 1
false_posR = false_pos/ (true_neg + false_pos)
balance = (false_neg + true_pos)/(true_neg + false_pos)
pop = false_neg + true_pos +true_neg + false_pos
self.metrics['balance'] = balance
self.metrics['False Positive Rate'] = false_posR
self.metrics['pop'] = pop
def build_conf_matrix(self, splits, X=None, y=None, verbose=False):
if not X or not y:
X = self.data['features']
y = self.data['labels']
X_train, y_train, X_test, y_test = split_data(X, y, splits)
## Want to step through the X_test and match up with y_test manually
cm = ConfusionMatrix()
for i, X_i in enumerate(X_test):
pred = self.predict(X_i.reshape(1, -1))
true = int(y_test[i])
pred = int(pred)
cm.add_value(true, pred)
matrix = cm.print_matrix()
return matrix