def update_or_delete_task(id):
service = DataService(current_app.config["app_context"])
request_data = request.get_json()
try:
data_exist = service.check_exist_by_id(id)
if request.method == "PUT":
status = request_data.get("status")
name = request_data.get("name")
_id = request_data.get("id")
if int(_id) != id:
return json.jsonify({"error": "id is not match."}), 400
if status == None or name == None:
return json.jsonify({"error": "missing request"}), 400
if not isinstance(status, int) or not isinstance(name, str):
return json.jsonify({"error": "request type wrong"}), 400
result = service.update_task(id=_id, status=status, name=name)
return json.jsonify(result), 200
if request.method == "DELETE":
result = service.delete_task(id)
return json.jsonify({"success": "delete success"}), 200
except RuntimeError as exp:
return json.jsonify({"error": str(exp)}), 400
except Exception as e:
return json.jsonify({"error": "server error"}), 500
def run(self):
data = DataService.load_csv("data/blood_pressure_cengage.csv")
feature_data = data[:, :-1]
labels = data[:, -1]
self.linear_regression_learner = LinearRegressionLearner(
data, learning_rate=0.001)
# PlotService.plot3d_scatter(data, labels=['Age', 'Weight', 'BP'], title="Blood Pressure for Age and Weight.")
normalized_data = DataService.normalize(data, method='min-max')
normalized_feature_data = normalized_data[:, :-1]
normalized_labels = normalized_data[:, -1]
# PlotService.plot3d_scatter(normalized_data, labels=['Age', 'Weight', 'BP'],
# title="BP for Age and Weight. Min-Max normalized.")
self.train_with_gradient_descent(feature_data, labels,
normalized_feature_data,
normalized_labels)
self.test(data)
self.train_with_normal_equation(feature_data, labels,
normalized_feature_data,
normalized_labels)
self.test(data)
def list_tasks():
service = DataService(current_app.config["app_context"])
try:
result = service.find_all_data()
return json.jsonify(result=result), 200
except Exception as e:
return json.jsonify({"error": "server error"}), 500
def new_task():
service = DataService(current_app.config["app_context"])
request_data = request.get_json()
name = request_data.get("name", None)
if not name:
return json.jsonify({"error": "key error: name"}), 400
try:
result = service.create_task(name)
return json.jsonify(result=result), 201
except Exception as e:
return json.jsonify({"error": "server error"}), 500
def run(self, k=2):
data = DataService.load_csv("data/wdbc.data")
# column 1 is the id, column 2 is the label, the rest are features
labels_data = data[:, 1]
self.kmeans_learner = KMeansLearner()
normalized_data = DataService.normalize(
data, method=self.normalization_method)
normalized_feature_data = normalized_data[:, 2:]
self.kmeans_learner.train(normalized_feature_data, k=k)
cluster_assignments = self.kmeans_learner.classify(
normalized_feature_data)
self.report_service.report(cluster_assignments, labels_data, k)
def run(self):
data = DataService.load_csv("data/wdbc.data")
# column 1 is the id, column 2 is the label, the rest are features
feature_data = data[:, 2:]
labels = data[:, 1]
config = NeuralNetConfig(num_inputs=feature_data.shape[1],
num_nodes_hidden_layers=[16],
num_outputs=1,
learning_rate=0.01,
epochs=800)
self.neural_net_learner = NeuralNetLearner(config)
normalized_data = DataService.normalize(data, method='z')
normalized_feature_data = normalized_data[:, 2:]
self.train_with_gradient_descent(data, labels, normalized_feature_data)
def run(self):
data = DataService.load_csv("data/wdbc.data")
# column 1 is the id, column 2 is the label, the rest are features
labels_data = data[:, 1]
self.knn_learner = KnnLearner()
normalized_data = DataService.normalize(data, method='z')
normalized_feature_data = normalized_data[:, 2:]
self.knn_learner.train(normalized_feature_data, labels_data, k=5)
predictions = []
for test_record in normalized_feature_data:
prediction = self.knn_learner.predict(test_record)
predictions.append(prediction)
predictions = np.array(predictions)
self.report_service.report(data, predictions, labels_data, self.knn_learner)
def run(self, k=2):
data = DataService.load_csv("data/wdbc.data")
# column 1 is the id, column 2 is the label, the rest are features
labels = data[:, 1]
self.em_learner = EMLearner()
normalized_data = DataService.normalize(
data, method=self.normalization_method)
normalized_feature_data = normalized_data[:, 2:]
feature_data = data[:, 2:]
# feature_data = normalized_feature_data
self.em_learner.train(feature_data, labels, k=k)
record_k_gaussian_probabilites, cluster_assignments = self.em_learner.estimate_probabilities(
feature_data)
self.report_service.accuracy_stats_report(
data, record_k_gaussian_probabilites, cluster_assignments, labels)
self.report_service.report(cluster_assignments, labels,
self.em_learner.k)
def run(self):
data = DataService.load_csv("data/blood_pressure_cengage.csv")
feature_data = data[:, :-1]
labels = data[:, -1]
#add one for the bias
feature_data = np.insert(feature_data, 0, 1, axis=1)
features_tensor = torch.tensor(feature_data.astype(np.float32))
labels_tensor = torch.tensor(labels.astype(np.float32))
train_epochs = 500
self.linear_regression_learner.train(features_tensor,
labels_tensor,
epochs=train_epochs)
trained_predictions = self.linear_regression_learner.predict(
features_tensor)
print(
"|Age|Weight|Actual Blood Pressure|Predicted Blood Pressure|Loss|")
print(
"|:-------:|:---:|:--------------------:|:---------------------:|")
train_records = data[:, :-1]
predicted_labels = []
for record_index, prediction in enumerate(trained_predictions):
age = train_records[record_index, 0]
weight = train_records[record_index, 1]
label = labels[record_index]
predicted_label = np.rint(prediction.data.numpy()[0])
predicted_labels.append(predicted_label)
loss = np.abs(label - predicted_label)
print("|{0}|{1}|{2}|{3}|{4}|".format(age, weight, label,
predicted_label, loss))
PlotService.plot_line(x=np.arange(train_epochs),
y=self.linear_regression_learner.loss_history,
x_label="Torch Epochs",
y_label="Loss",
title="Absolute Loss per Epoch")
PlotService.plot3d_scatter_compare(train_records,
labels,
trained_predictions.data.numpy(),
labels=['Age', 'Weight', 'BP'],
title="Actual vs Projected")
def run(self, k=2):
data = DataService.load_csv("data/wdbc.data")
# column 1 is the id, column 2 is the label, the rest are features
feature_data = data[:, 2:]
labels_data = data[:, 1]
self.naive_bayes_learner = NaiveBayesLearner()
self.naive_bayes_learner.train(feature_data, labels_data)
predictions = []
for record_index, feature_record in enumerate(feature_data):
prediction = self.naive_bayes_learner.predict(feature_record)
predictions.append(prediction)
self.report_service.report(data, predictions, labels_data, self.naive_bayes_learner)
def run(self, k=2):
feature_data, labels = DataService.get_data("data/play_tennis.csv")
self.decision_tree_learner.train(feature_data, labels)
errors = 0
for test_index in range(labels.size):
prediction = self.decision_tree_learner.classify(
feature_data[test_index, :])
actual = labels[test_index]
if prediction != actual:
errors += 1
accuracy = (labels.size - errors) / labels.size
print("Accuracy: ", accuracy)
def run(self):
data = DataService.load_csv("data/blood_pressure_cengage.csv")
feature_data = data[:, :-1]
labels = data[:, -1].reshape(feature_data.shape[0], 1)
lin_reg = LinearRegression()
lin_reg_model = lin_reg.fit(feature_data, labels)
predictions = lin_reg_model.predict(feature_data)
ReportsService.print_results_table(data, predictions, labels)
PlotService.plot3d_scatter_compare(
feature_data,
labels,
predictions,
labels=['Age', 'Weight', 'BP'],
title="Actual vs Projected LinearRegressor")
def train_with_normal_equation(self, feature_data, labels_data,
normalized_feature_data, normalized_labels):
print("Normal Equation: ")
# Normal equation
feature_data_bias = np.insert(normalized_feature_data, 0, 1, axis=1)
x_trans_x = np.dot(np.transpose(feature_data_bias), feature_data_bias)
norm_equation_theta = np.dot(
np.dot(np.linalg.inv(x_trans_x), np.transpose(feature_data_bias)),
normalized_labels)
print(norm_equation_theta)
self.linear_regression_learner.theta = norm_equation_theta
normalized_predictions = self.linear_regression_learner.predict(
normalized_feature_data)
predictions = np.around(
DataService.denormalize_predictions(labels_data,
normalized_predictions,
method='min-max'))
cost = self.linear_regression_learner.calculate_cost(
normalized_predictions, normalized_labels)
print("Final Norm Equation Normalized Cost: ", cost)
x, y, z = self.build_model_plot_data(feature_data, predictions)
PlotService.plot3d_line(
x,
y,
z,
labels=['Age', 'Weight', 'BP'],
title="Normal Equation Model: Blood Pressure for Age and Weight.")
PlotService.plot3d_scatter_compare(
feature_data,
labels_data,
predictions,
labels=['Age', 'Weight', 'BP'],
title="Normal Equation Actual vs Projected")
def train_with_gradient_descent(self, feature_data, labels_data,
normalized_feature_data,
normalized_labels):
self.linear_regression_learner.train(normalized_feature_data,
normalized_labels)
normalized_predictions = self.linear_regression_learner.predict(
normalized_feature_data)
predictions = np.around(
DataService.denormalize_predictions(labels_data,
normalized_predictions,
method='min-max'))
PlotService.plot_line(x=range(
1,
len(self.linear_regression_learner.cost_history) + 1),
y=self.linear_regression_learner.cost_history,
x_label="Iteration",
y_label="Training Cost",
title="Training Learning Curve")
x, y, z = self.build_model_plot_data(feature_data, predictions)
PlotService.plot3d_line(
x,
y,
z,
labels=['Age', 'Weight', 'BP'],
title="Linear Model: Blood Pressure for Age and Weight.")
PlotService.plot3d_scatter_compare(feature_data,
labels_data,
predictions,
labels=['Age', 'Weight', 'BP'],
title="Actual vs Projected")
cost = self.linear_regression_learner.calculate_cost(
normalized_predictions, normalized_labels)
print("Final Normalized Cost: ", cost)
import tensorflow as tf
from service.data_service import DataService
import numpy as np
from service.plot_service import PlotService
data = DataService.load_csv("data/blood_pressure_cengage.csv")
def get_data():
feature_data = data[:, :-1]
# add 1 at the beginning for the bias
feature_data = np.insert(feature_data, 0, 1, axis=1)
labels = data[:, -1]
return {"Age": feature_data[:, 0], "Weight": feature_data[:, 1]}, labels
def get_feature_data():
feature_data, _ = get_data()
return feature_data
def get_labels():
_, labels = get_data()
return labels
age_column = tf.feature_column.numeric_column('Age')
weight_column = tf.feature_column.numeric_column('Weight')
lin_reg = tf.estimator.LinearRegressor(feature_columns=[age_column, weight_column])
steps_history = []
def test(self, data):
x_mins = np.amin(data, axis=0)
x_maxs = np.amax(data, axis=0)
min_age = x_mins[0]
max_age = x_maxs[0]
min_weight = x_mins[1]
max_weight = x_maxs[1]
min_blood_pressure = x_mins[2]
max_blood_pressure = x_maxs[2]
x_means = np.mean(data, axis=0)
mean_age = x_means[0]
mean_weight = x_means[1]
mean_blood_pressure = x_means[2]
x_stds = np.std(data, axis=0)
std_age = x_stds[0]
std_weight = x_stds[1]
std_blood_pressure = x_stds[2]
while True:
age = input("Enter Age or type quit to exit: ")
if age == 'quit':
break
weight = input("Enter Weight or type quit to exit: ")
if weight == 'quit':
break
try:
age = int(age)
weight = int(weight)
except ValueError:
print("Age and Weight should be integers.", age, weight)
break
test_age = DataService.normalize_single(
age,
min=min_age,
max=max_age,
mean=mean_age,
std=std_age,
method=self.normalization_method)
test_weight = DataService.normalize_single(
weight,
min=min_weight,
max=max_weight,
mean=mean_weight,
std=std_weight,
method=self.normalization_method)
current_theta_normalized_projection = \
self.linear_regression_learner.predict(np.array([test_age, test_weight]).reshape(1, 2))
print(
"Current Theta Projection: ",
DataService.denormalize_single(
current_theta_normalized_projection,
method=self.normalization_method,
min=min_blood_pressure,
max=max_blood_pressure,
mean=mean_blood_pressure,
std=std_blood_pressure))
# best documented theta min-max norm: -0.04973713 0.7198039 0.33875262
# best documented theta zscore norm: -0.03447741 0.70838301 0.32258052
best_theta = np.array([-0.03447741, 0.70838301,
0.32258052]).reshape((1, 3))
if self.normalization_method == 'min-max':
best_theta = np.array([-0.04973713, 0.7198039,
0.33875262]).reshape((1, 3))
best_theta_normalized_projection = self.linear_regression_learner\
.predict_for_theta(np.array([test_age, test_weight]).reshape(1, 2), best_theta)
print(
"Best Theta Projection: ",
DataService.denormalize_single(
best_theta_normalized_projection,
method=self.normalization_method,
min=min_blood_pressure,
max=max_blood_pressure,
mean=mean_blood_pressure,
std=std_blood_pressure))
def setUp(self):
self.service = DataService(app_context)
class ServiceTestCase(unittest.TestCase):
task_data = None
def setUp(self):
self.service = DataService(app_context)
def test_0100_find_all_data_success(self):
find_all_patch = patch.object(DataDao,
'find_all',
side_effect=self.mock_find_all)
find_all_patch.start()
return_data = self.service.find_all_data()
assert return_data == [{
"id": 1,
"name": "name",
"status": 0
}, {
"id": 2,
"name": "name2",
"status": 1
}]
find_all_patch.stop()
def test_0200_check_exist_by_id_success(self):
"""
Test case:
return_data must be true
"""
check_exist_by_id_patch = patch.object(
DataDao,
'check_exist_by_id',
side_effect=self.mock_check_exist_by_id)
check_exist_by_id_patch.start()
id = 1
return_data = self.service.check_exist_by_id(id)
assert return_data == True
check_exist_by_id_patch.stop()
def test_0201_check_exist_by_id_fail(self):
"""
Test case:
function will raise RuntimeError
"""
self.assertRaisesRegex(RuntimeError, "id is not found in database",
self.check_exist_by_id_fail)
def check_exist_by_id_fail(self):
check_exist_by_id_patch = patch.object(
DataDao,
'check_exist_by_id',
side_effect=self.mock_check_exist_by_id)
check_exist_by_id_patch.start()
id = 2
self.service.check_exist_by_id(id)
check_exist_by_id_patch.stop()
def test_0300_create_task_success(self):
"""
Test case:
create task will return data
"""
name = "test"
insert_data_patch = patch.object(DataDao,
'insert_data',
side_effect=self.mock_insert_data)
get_by_id_patch = patch.object(DataDao,
'get_by_id',
side_effect=self.mock_get_by_id)
get_by_id_patch.start()
insert_data_patch.start()
return_data = self.service.create_task(name)
insert_data_patch.stop()
get_by_id_patch.stop()
assert return_data == {"id": 1, "name": name, "status": 0}
def test_0400_update_task_success(self):
"""
Test case:
update task will return data
"""
_id = 1
status = 1
name = "test2"
update_data_patch = patch.object(DataDao,
'update_data',
side_effect=self.mock_update_data)
get_by_id_patch = patch.object(DataDao,
'get_by_id',
side_effect=self.mock_get_by_id)
get_by_id_patch.start()
update_data_patch.start()
return_data = self.service.update_task(_id, status=status, name=name)
update_data_patch.stop()
get_by_id_patch.stop()
assert return_data == {"id": 1, "name": "test2", "status": 1}
def test_0500_delete_task_success(self):
"""
Test case:
delete task success
"""
_id = 1
delete_data_patch = patch.object(DataDao,
'delete_data',
side_effect=self.mock_delete_data)
delete_data_patch.start()
return_data = self.service.delete_task(_id)
delete_data_patch.stop()
assert return_data == "success"
def test_0501_delete_task_fail(self):
"""
Test case:
delete task fail
"""
self.assertRaisesRegex(RuntimeError, "id is not found in database",
self.delete_task_fail)
def delete_task_fail(self):
_id = 2
delete_data_patch = patch.object(DataDao,
'delete_data',
side_effect=self.mock_delete_data)
delete_data_patch.start()
return_data = self.service.delete_task(_id)
delete_data_patch.stop()
def mock_find_all(self, *args, **kwargs):
return [{
"id": 1,
"name": "name",
"status": 0
}, {
"id": 2,
"name": "name2",
"status": 1
}]
def mock_check_exist_by_id(self, id):
if id == 1:
return True
else:
return False
def mock_insert_data(self, name):
self.__class__.task_data = {"id": 1, "name": name, "status": 0}
return 1
def mock_get_by_id(self, id):
if id == 1:
return self.task_data
def mock_update_data(self, id, **kwargs):
self.__class__.task_data = {
"id": id,
"name": kwargs["name"],
"status": kwargs["status"]
}
def mock_delete_data(self, id):
if id == 1:
return True
else:
return False
