def show_improved_model_performances(dframe):
print('\nCalculating improved models performances:')
train, test = preprocess.get_train_test_split(dframe)
test_x = preprocess.get_featues(test)
test_y = preprocess.get_label(test)
show_models_performances([
get_model(RandomForestModel()),
get_model(LogisticRegressionModel()),
get_model(SVM()),
get_model_keras()
], test_x, test_y)
def show_benchmark_model_performance():
print('Calculating benchmark model performance:')
dframe = preprocess.read_processed_data(constants.RAW_DATA_PICKLE)
train, test = preprocess.get_train_test_split(dframe)
test_x = preprocess.get_featues(test)
test_y = preprocess.get_label(test)
model = SVM()
model_file = os.path.join(constants.PROJ_ROOT, 'models', 'benchmark',
model.name + '.model')
model.load(model_file)
show_models_performances([model],
test_x,
test_y,
roc_file=constants.BENCHMARK_ROC_PATH)
def train(self, train, validation):
X = get_featues(train)
y = get_label(train)
xv = get_featues(validation)
yv = get_label(validation)
self.clf.fit(X, y, epochs=500, verbose=0, validation_data=(xv, yv))
def train(self, dframe):
X = get_featues(dframe)
y = get_label(dframe)
self.clf.fit(X, y)