def test_regressor_toy(regressor):
X = [
{'outlook': 'rain', 'temperature': 'hot', 'humidity': 'high', 'wind': False},
{'outlook': 'rain', 'temperature': 'hot', 'humidity': 'high', 'wind': True},
{'outlook': 'overcast', 'temperature': 'hot', 'humidity': 'high', 'wind': False},
{'outlook': 'sunny', 'temperature':'mild', 'humidity': 'high', 'wind': False},
{'outlook': 'sunny', 'temperature':'cool', 'humidity': 'normal', 'wind': False},
{'outlook': 'sunny', 'temperature':'cool', 'humidity': 'normal', 'wind': True},
{'outlook': 'overcast', 'temperature':'cool', 'humidity': 'normal', 'wind': True},
{'outlook': 'rain', 'temperature':'mild', 'humidity': 'high', 'wind': False},
{'outlook': 'rain', 'temperature':'cool', 'humidity': 'normal', 'wind': False},
{'outlook': 'sunny', 'temperature':'mild', 'humidity': 'normal', 'wind': False},
{'outlook': 'rain', 'temperature':'mild', 'humidity': 'normal', 'wind': True},
{'outlook': 'overcast', 'temperature':'mild', 'humidity': 'high', 'wind': True},
{'outlook': 'overcast', 'temperature':'hot', 'humidity': 'normal', 'wind': False},
{'outlook': 'sunny', 'temperature':'mild', 'humidity': 'high', 'wind': True},
]
Y = np.array([26,30,48,46,62,23,43,36,38,48,48,62,44,30])
vectorizer = DictVectorizer()
data = vectorizer.fit_transform(X)
print(vectorizer.feature_to_index)
print(vectorizer.index_to_feature_type)
regressor.set_index_to_feature_type(index_to_feature_type=vectorizer.index_to_feature_type)
regressor.fit(data, Y)
#tree.print_tree()
#x = vectorizer.transform({'outlook': 'rain', 'temperature':'mild',
# 'humidity': 'normal', 'wind': False})
#print(f"predicting: ['rain', 'mild', 'normal', False] = {x}")
cross_validation(regressor, data, Y, task_type='regression', num_folds =len(Y))
def test_regressor_boston_house(regressor):
# this actually takes many seconds, even though the data is small
d = load_boston()
index_to_feature_type = defaultdict(lambda: 'numerical')
regressor.set_index_to_feature_type(index_to_feature_type)
#regressor.fit(d.data, d.target)
cross_validation(regressor, d.data, d.target, task_type='regression', num_folds=5)
def test_classifier_breast_cancer(classifier):
# this actually takes many seconds, even though the data is small
d = load_breast_cancer()
index_to_feature_type = defaultdict(lambda: 'numerical')
classifier.set_index_to_feature_type(index_to_feature_type)
#classifier.fit(d.data, d.target)
print(d.data.shape)
cross_validation(classifier, d.data, d.target, task_type='classification', num_folds=3)
def test_decision_tree_regressor_boston_house():
# this actually takes many seconds, even though the data is small
d = load_boston()
print(d)
tree = DecisionTreeRegressor(max_depth=3)
index_to_feature_type = defaultdict(lambda: 'numerical')
tree.set_index_to_feature_type(index_to_feature_type)
tree.fit(d.data, d.target)
tree.print_tree()
cross_validation(tree,
d.data,
d.target,
task_type='regression',
num_folds=5)
def test_random_forest_classifier_breast_cancer():
# this actually takes many seconds, even though the data is small
d = load_breast_cancer()
forest = RandomForestClassifier(num_trees=50,
max_depth=None,
num_features_to_sample_from=(int(
d.data.shape[1] / 3)))
index_to_feature_type = defaultdict(lambda: 'numerical')
forest.set_index_to_feature_type(index_to_feature_type)
#forest.fit(d.data, d.target)
#forest.print_tree()
cross_validation(forest,
d.data,
d.target,
num_folds=5,
task_type='classification')
def test_linear_regressor_diabetes():
d = load_diabetes()
print(d)
regressor = LinearRegressor(max_iters=5000)
#regressor.fit(d.data, d.target)
#print(regressor.b)
#print(regressor.thetas)
regression = LinearRegression()
#regression.fit(d.data, d.target)
#print(regression.intercept_)
#print(regression.coef_)
cross_validation(regressor,
d.data,
d.target,
task_type='regression',
num_folds=10)
def test_classifier_toy(classifier):
# decision tree people really want to know if they should play tennis
X = [
{'outlook': 'sunny', 'temperature': 'hot', 'humidity': 'high', 'wind': 'weak'},
{'outlook': 'sunny', 'temperature': 'hot', 'humidity': 'high', 'wind': 'strong'},
{'outlook': 'overcast', 'temperature': 'hot', 'humidity': 'high', 'wind': 'weak'},
{'outlook': 'rain', 'temperature':'mild', 'humidity': 'high', 'wind': 'weak'},
{'outlook': 'rain', 'temperature':'cool', 'humidity': 'normal', 'wind': 'weak'},
{'outlook': 'rain', 'temperature':'cool', 'humidity': 'normal', 'wind': 'strong'},
{'outlook': 'overcast', 'temperature':'cool', 'humidity': 'normal', 'wind': 'strong'},
{'outlook': 'sunny', 'temperature':'mild', 'humidity': 'high', 'wind': 'weak'},
{'outlook': 'sunny', 'temperature':'cool', 'humidity': 'normal', 'wind': 'weak'},
{'outlook': 'rain', 'temperature':'mild', 'humidity': 'normal', 'wind': 'weak'},
{'outlook': 'sunny', 'temperature':'mild', 'humidity': 'normal', 'wind': 'strong'},
{'outlook': 'overcast', 'temperature':'mild', 'humidity': 'high', 'wind': 'strong'},
{'outlook': 'overcast', 'temperature':'hot', 'humidity': 'normal', 'wind': 'weak'},
{'outlook': 'rain', 'temperature':'mild', 'humidity': 'high', 'wind': 'strong'},
]
Y = np.array([0,0,1,1,1,0,1,0,1,1,1,1,1,0])
vectorizer = DictVectorizer()
data = vectorizer.fit_transform(X)
print(vectorizer.feature_to_index)
print(vectorizer.index_to_feature_type)
try:
classifier.set_index_to_feature_type(index_to_feature_type=vectorizer.index_to_feature_type)
except:
# kinda gross. i wanna refactor this to not need the vectorizer to pass in this info.
# just let the fit methods figure this out. less work then for the general API
pass
#classifier.fit(data, Y)
#classifier.print_tree()
x = vectorizer.transform({'outlook': 'rain', 'temperature':'mild',
'humidity': 'normal', 'wind': 'weak'})
#print(f"predicting: ['rain', 'mild', 'normal', 'weak'] = {x}")
#print(classifier.predict(x))
cross_validation(classifier, data, Y, task_type='classification', num_folds=3)