def test_feature_importance_regression():
"""Test that Gini importance is calculated correctly.
This test follows the example from [1]_ (pg. 373).
.. [1] Friedman, J., Hastie, T., & Tibshirani, R. (2001). The elements
of statistical learning. New York: Springer series in statistics.
"""
california = fetch_california_housing()
X, y = california.data, california.target
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)
reg = GradientBoostingRegressor(loss='huber', learning_rate=0.1,
max_leaf_nodes=6, n_estimators=100,
random_state=0)
reg.fit(X_train, y_train)
sorted_idx = np.argsort(reg.feature_importances_)[::-1]
sorted_features = [california.feature_names[s] for s in sorted_idx]
# The most important feature is the median income by far.
assert sorted_features[0] == 'MedInc'
# The three subsequent features are the following. Their relative ordering
# might change a bit depending on the randomness of the trees and the
# train / test split.
assert set(sorted_features[1:4]) == {'Longitude', 'AveOccup', 'Latitude'}
def test_krige_housing():
try:
housing = fetch_california_housing()
except PermissionError:
# This can raise permission error on Appveyor
pytest.skip('Failed to load california housing dataset')
# take only first 1000
p = housing['data'][:1000, :-2]
x = housing['data'][:1000, -2:]
target = housing['target'][:1000]
p_train, p_test, y_train, y_test, x_train, x_test = \
train_test_split(p, target, x, train_size=0.7,
random_state=10)
for ml_model, krige_method in _methods():
reg_kr_model = RegressionKriging(regression_model=ml_model,
method=krige_method,
n_closest_points=2)
reg_kr_model.fit(p_train, x_train, y_train)
if krige_method == 'ordinary':
assert reg_kr_model.score(p_test, x_test, y_test) > 0.5
else:
assert reg_kr_model.score(p_test, x_test, y_test) > 0.0
def load_data_target(name):
"""
Loads data and target given the name of the dataset.
"""
if name == "Boston":
data = load_boston()
elif name == "Housing":
data = fetch_california_housing()
dataset_size = 1000 # this is necessary so that SVR does not slow down too much
data["data"] = data["data"][:dataset_size]
data["target"] =data["target"][:dataset_size]
elif name == "digits":
data = load_digits()
elif name == "Climate Model Crashes":
try:
data = fetch_mldata("climate-model-simulation-crashes")
except HTTPError as e:
url = "https://archive.ics.uci.edu/ml/machine-learning-databases/00252/pop_failures.dat"
data = urlopen(url).read().split('\n')[1:]
data = [[float(v) for v in d.split()] for d in data]
samples = np.array(data)
data = dict()
data["data"] = samples[:, :-1]
data["target"] = np.array(samples[:, -1], dtype=np.int)
else:
raise ValueError("dataset not supported.")
return data["data"], data["target"]
def getCaliforniaHousingData(housingFILE):
print("\ngetCaliforniaHousingData():")
### ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ###
if os.path.isfile(path = housingFILE) == False:
print("downloading California housing data set from Internet ...")
from sklearn.datasets import fetch_california_housing
with open(file = housingFILE, mode = 'wb') as myFile:
pickle.dump(obj = fetch_california_housing(), file = myFile)
else:
print("loading California housing data set from drive ...")
### ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ###
with open(file = housingFILE, mode = 'rb') as myFile:
housing = pickle.load(myFile)
### ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ###
housingData = housing.data
housingTarget = housing.target
### ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ###
return( housingData, housingTarget )
from sklearn.datasets import fetch_california_housing
from sklearn.impute import SimpleImputer
from sklearn.impute import IterativeImputer
from sklearn.linear_model import BayesianRidge
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import ExtraTreesRegressor
from sklearn.neighbors import KNeighborsRegressor
from sklearn.pipeline import make_pipeline
from sklearn.model_selection import cross_val_score
N_SPLITS = 5
rng = np.random.RandomState(0)
X_full, y_full = fetch_california_housing(return_X_y=True)
# ~2k samples is enough for the purpose of the example.
# Remove the following two lines for a slower run with different error bars.
X_full = X_full[::10]
y_full = y_full[::10]
n_samples, n_features = X_full.shape
# Estimate the score on the entire dataset, with no missing values
br_estimator = BayesianRidge()
score_full_data = pd.DataFrame(
cross_val_score(
br_estimator, X_full, y_full, scoring='neg_mean_squared_error',
cv=N_SPLITS
),
columns=['Full Data']
)
#
# In the previous notebook, we presented the general cross-validation framework
# and how to assess if a predictive model is underfiting, overfitting, or
# generalizing. Besides these aspects, it is also important to understand how
# the different errors are influenced by the number of samples available.
#
# In this notebook, we will show this aspect by looking a the variability of
# the different errors.
#
# Let's first load the data and create the same model as in the previous
# notebook.
# %%
from sklearn.datasets import fetch_california_housing
housing = fetch_california_housing(as_frame=True)
data, target = housing.data, housing.target
target *= 100 # rescale the target in k$
# %% [markdown]
# ```{note}
# If you want a deeper overview regarding this dataset, you can refer to the
# Appendix - Datasets description section at the end of this MOOC.
# ```
# %%
from sklearn.tree import DecisionTreeRegressor
regressor = DecisionTreeRegressor()
# %% [markdown]
You should have received a copy of the GNU General Public License along with this program. If not, see <http://www.gnu.org/licenses/>. Copyright Brian Dolhansky 2014 [email protected] """ import numpy as np from data_utils import split_train_test, RMSE from linear_regression import LinearRegression from sklearn import preprocessing from sklearn.datasets import fetch_california_housing print "Loading data..." housing = fetch_california_housing(data_home='/home/bdol/data') train_data, test_data, train_target, test_target = split_train_test( housing.data, housing.target ) # Normalize the data train_data = preprocessing.scale(train_data) test_data = preprocessing.scale(test_data) # Append bias feature train_data = np.hstack((train_data, np.ones((train_data.shape[0], 1), dtype=train_data.dtype))) test_data = np.hstack((test_data, np.ones((test_data.shape[0], 1), dtype=test_data.dtype))) train_target = train_target[:, None]
You should have received a copy of the GNU General Public License along with this program. If not, see <http://www.gnu.org/licenses/>. Copyright Brian Dolhansky 2014 [email protected] """ import numpy as np from data_utils import split_train_test, RMSE from linear_regression import LinearRegression from sklearn import preprocessing from sklearn.datasets import fetch_california_housing print "Loading data..." housing = fetch_california_housing(data_home='/home/bdol/data') train_data, test_data, train_target, test_target = split_train_test( housing.data, housing.target) # Normalize the data train_data = preprocessing.scale(train_data) test_data = preprocessing.scale(test_data) # Append bias feature train_data = np.hstack( (train_data, np.ones((train_data.shape[0], 1), dtype=train_data.dtype))) test_data = np.hstack( (test_data, np.ones((test_data.shape[0], 1), dtype=test_data.dtype))) train_target = train_target[:, None] test_target = test_target[:, None]
# cross-validation could be used. # # This exercise will make you implement the same search but using the class # `GridSearchCV`. # # First, we will: # # * load the california housing dataset; # * split the data into a training and testing set; # * create a machine learning pipeline composed of a standard scaler to # normalize the data, and a ridge regression as a linear model. # %% from sklearn.datasets import fetch_california_housing X, y = fetch_california_housing(as_frame=True, return_X_y=True) X.head() # %% from sklearn.model_selection import train_test_split X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0) # %% from sklearn.pipeline import make_pipeline from sklearn.preprocessing import PolynomialFeatures, StandardScaler from sklearn.linear_model import Ridge ridge = make_pipeline(StandardScaler(), Ridge()) # %% [markdown]
def test_rf_regression(datatype, split_algo, mode, column_info, max_features,
rows_sample):
ncols, n_info = column_info
use_handle = True
if mode == 'unit':
X, y = make_regression(n_samples=500,
n_features=ncols,
n_informative=n_info,
random_state=123)
elif mode == 'quality':
X, y = fetch_california_housing(return_X_y=True)
else:
X, y = make_regression(n_samples=100000,
n_features=ncols,
n_informative=n_info,
random_state=123)
X = X.astype(datatype)
y = y.astype(datatype)
X_train, X_test, y_train, y_test = train_test_split(X,
y,
train_size=0.8,
random_state=0)
# Create a handle for the cuml model
handle, stream = get_handle(use_handle, n_streams=1)
# Initialize and fit using cuML's random forest regression model
cuml_model = curfr(max_features=max_features,
rows_sample=rows_sample,
n_bins=16,
split_algo=split_algo,
split_criterion=2,
min_rows_per_node=2,
seed=123,
n_streams=1,
n_estimators=50,
handle=handle,
max_leaves=-1,
max_depth=16,
accuracy_metric='mse')
cuml_model.fit(X_train, y_train)
# predict using FIL
fil_preds = cuml_model.predict(X_test, predict_model="GPU")
cu_preds = cuml_model.predict(X_test, predict_model="CPU")
cu_r2 = r2_score(y_test, cu_preds, convert_dtype=datatype)
fil_r2 = r2_score(y_test, fil_preds, convert_dtype=datatype)
# Initialize, fit and predict using
# sklearn's random forest regression model
if mode != "stress":
sk_model = skrfr(n_estimators=50,
max_depth=16,
min_samples_split=2,
max_features=max_features,
random_state=10)
sk_model.fit(X_train, y_train)
sk_predict = sk_model.predict(X_test)
sk_r2 = r2_score(y_test, sk_predict, convert_dtype=datatype)
assert fil_r2 >= (sk_r2 - 0.07)
assert fil_r2 >= (cu_r2 - 0.02)
import tensorflow as tf
import numpy as np
from sklearn.datasets import fetch_california_housing
# 获取房价数据集
housing = fetch_california_housing(data_home='./scikit_learn_data',
download_if_missing=True)
# m 行， n 列
m, n = housing.data.shape
feature_names = housing.feature_names
# 20640 8
# ['MedInc', 'HouseAge', 'AveRooms', 'AveBedrms', 'Population', 'AveOccup', 'Latitude', 'Longitude']
print(m, n)
print(feature_names)
print(housing.data[0:3], housing.target[0:3], type(housing.target[0:3]))
# np.c_ 将两个矩阵拼接在一起
housing_data_plus_bias = np.c_[np.ones((m, 1)), housing.data]
print(housing_data_plus_bias.shape)
# 创建两个 Tensorflow 的常量节点 X 和 y ， 去持有数据和标签
X = tf.constant(housing_data_plus_bias, dtype=tf.float32, name='X')
# X = tf.constant(housing.data, dtype=tf.float32, name='X')
# 行向量转为列向量
y = tf.constant(housing.target.reshape(-1, 1), dtype=tf.float32, name='y')
# 使用 Tensorflow 框架提供的矩阵操作操作求 theta
XT = tf.transpose(X)
# 解析解一步计算出最优解
theta = tf.matmul(tf.matmul(tf.matrix_inverse(tf.matmul(XT, X)), XT), y)
# Tensor("MatMul_2:0", shape=(9, 1), dtype=float32)
print(theta)
import time
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument("--tabularpath", type=str, default='',
nargs="?", help="Path of Tabular Data, i.e./.../data.csv")
# Hyper parameters for conversion
parser.add_argument("--num_quantile", type=int, default=2, nargs="?",
help="q param in https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.qcut.html")
parser.add_argument("--min_unique_val_per_column", type=int, default=2, nargs="?",
help="Apply Quantile-based discretization function on those columns having at least such "
"unique values.")
args = parser.parse_args()
# DASK can be applied.
print('Tabular data is being read')
try:
df = pd.read_csv(args.tabularpath)
except FileNotFoundError:
print('File not found, We will use california housing dataset from sklearn')
from sklearn import datasets
X, y = datasets.fetch_california_housing(return_X_y=True)
df = pd.DataFrame(X)
print('Original Tabular data: {0} by {1}'.format(*df.shape))
print('Quantisation starts')
X_transformed = QCUT(min_unique_val_per_column=args.min_unique_val_per_column,
num_quantile=args.num_quantile).transform(df)
X_transformed.index = 'Event_' + X_transformed.index.astype(str)
print('Graph data being generated')
kg = GraphGenerator().transform(X_transformed)
def california_housing_dataset(batch_size, device):
"""
This named constructor builds a DataSet from the California Housing dataset.
"""
from sklearn.datasets import fetch_california_housing
cal_housing = fetch_california_housing()
non_outliers = np.ones(cal_housing.data.shape[0], dtype=bool)
for idx in range(cal_housing.data.shape[1]):
column = cal_housing.data[:, idx]
cutoffs = np.percentile(column, (1.0, 99.0))
non_outliers = np.logical_and(
non_outliers,
np.logical_and(column > cutoffs[0], column < cutoffs[1]))
# cutoffs = np.percentile(cal_housing.target, (1.0, 99.0))
# non_outliers = np.logical_and(
# non_outliers, np.logical_and(cal_housing.target > cutoffs[0],
# cal_housing.target < cutoffs[1])
# )
cal_housing.data = cal_housing.data[non_outliers]
cal_housing.target = cal_housing.target[non_outliers]
x_train, x_test, y_train, y_test = train_test_split(cal_housing.data,
cal_housing.target,
test_size=0.2,
random_state=0)
non_outliers = np.ones(x_test.shape[0], dtype=bool)
for idx in range(x_test.shape[1]):
column = x_test[:, idx]
cutoffs = np.percentile(column, (5.0, 95.0))
non_outliers = np.logical_and(
non_outliers,
np.logical_and(column > cutoffs[0], column < cutoffs[1]))
x_test = x_test[non_outliers]
y_test = y_test[non_outliers]
x_dimensions = cal_housing.feature_names
train_size = x_train.shape[0]
test_size = x_test.shape[0]
y_train = y_train[..., np.newaxis]
y_test = y_test[..., np.newaxis]
params_desc = "train size: {}/test size: {}".format(
train_size, test_size)
return DataSets(
x_train,
y_train,
x_test,
y_test,
x_dimensions,
"Price",
batch_size,
"california housing dataset",
params_desc,
device,
)
import tensorflow as tf import numpy as np from sklearn.datasets import fetch_california_housing # 立刻下载数据集 housing = fetch_california_housing(download_if_missing=True) # 获得X数据行数和列数 m, n = housing.data.shape print(m, n) print(housing.data, housing.target) print(housing.feature_names) # 这里添加一个额外的bias输入特征(x0=1)到所有的训练数据上面，因为使用的numpy所以会立即执行 housing_data_plus_bias = np.c_[np.ones((m, 1)), housing.data] # 创建两个TensorFlow常量节点X和y，去持有数据和标签 X = tf.constant(housing_data_plus_bias, dtype=tf.float32, name='X') y = tf.constant(housing.target.reshape(-1, 1), dtype=tf.float32, name='y') # 使用一些TensorFlow框架提供的矩阵操作去求theta XT = tf.transpose(X) # 解析解一步计算出最优解 theta = tf.matmul(tf.matmul(tf.matrix_inverse(tf.matmul(XT, X)), XT), y) print(theta) # with tf.Session() as sess: # theta_value = theta.eval() # sess.run(theta) # # print(theta_value)
import tensorflow as tf
import numpy as np
from sklearn.datasets import fetch_california_housing
from sklearn.preprocessing import StandardScaler
#tensorflow多元线性回归梯度下降法求解
n_epochs = 10000
learning_rate = 0.01
housing = fetch_california_housing(data_home="D:/sklearn_data",
download_if_missing=True)
m, n = housing.data.shape
housing_data_plus_bias = np.c_[np.ones((m, 1)), housing.data]
# 可以使用TensorFlow或者Numpy或者sklearn的StandardScaler去进行归一化
# StandardScaler默认就做了方差归一化，和均值归一化，这两个归一化的目的都是为了更快的进行梯度下降
# 你如何构建你的训练集，你训练除了的模型，就具备什么样的功能！
scaler = StandardScaler().fit(housing_data_plus_bias)
scaled_housing_data_plus_bias = scaler.transform(housing_data_plus_bias)
X = tf.constant(scaled_housing_data_plus_bias, dtype=tf.float32, name='X')
y = tf.constant(housing.target.reshape(-1, 1), dtype=tf.float32, name='y')
# random_uniform函数创建图里一个节点包含随机数值，给定它的形状和取值范围，就像numpy里面rand()函数
theta = tf.Variable(tf.random_uniform([n + 1, 1], -1.0, 1.0), name='theta')
y_pred = tf.matmul(X, theta, name="predictions")
error = y_pred - y
mse = tf.reduce_mean(tf.square(error), name="mse")
# 梯度的公式：(y_pred - y) * xj
gradients = 2 / m * tf.matmul(tf.transpose(X), error)
# 赋值函数对于BGD来说就是 theta_new = theta - (learning_rate * gradients)
training_op = tf.assign(theta, theta - learning_rate * gradients)
import tensorflow as tf
import numpy as np
from sklearn.datasets import fetch_california_housing
housing = fetch_california_housing(data_home="G:\ML\dataset\scikit_learn_data",
download_if_missing=True)
m, n = housing.data.shape
print(m, n)
print(housing.target.shape)
print(housing.target)
#print(housing.data, housing.target)
print(housing.feature_names)
housing_data_plus_bias = np.c_[np.ones((m, 1)), housing.data]
y_pre = housing.target.reshape(-1, 1)
print(y_pre.shape)
X = tf.constant(housing_data_plus_bias, dtype=tf.float32, name='X')
Y = tf.constant(y_pre, dtype=tf.float32, name='y')
XT = tf.transpose(X)
theta = tf.matmul(tf.matmul(tf.matrix_inverse(tf.matmul(XT, X)), XT), Y)
with tf.Session() as sess:
theta_value = theta.eval()
print(theta_value)
from sklearn.datasets import fetch_california_housing
cali = fetch_california_housing() #bunch object
# print(cali.DESCR)
print(cali.data.shape)
print(cali.target.shape)
print(cali.feature_names)
import pandas as pd
pd.set_option("precision", 4)
pd.set_option("max_columns", 9) #display up to 9 columns in DataFrame
pd.set_option("display.width", None) # auto-detect the display width
cali_df = pd.DataFrame(cali.data, columns=cali.feature_names)
cali_df["MedHouseValue"] = pd.Series(cali.target)
print(cali_df.head())
sample_df = cali_df.sample(frac=0.1, random_state=17)
import matplotlib.pyplot as plt
import seaborn as sns
sns.set(font_scale=2)
sns.set_style("whitegrid")
for feature in cali.feature_names:
plt.figure(figsize=(8, 4.5)) # 8"-by-4.5" figure
sns.scatterplot(
data=sample_df,
def main():
assert not tf.executing_eagerly()
now = datetime.utcnow().strftime('%Y%m%d%H%M%S')
root_logdir = 'tf_logs'
logdir = '{}/run-{}/'.format(root_logdir, now)
housing = fetch_california_housing()
m, n = housing.data.shape
housing_data_plus_bias = np.c_[np.ones((m, 1)), housing.data]
scaler = MinMaxScaler()
housing_data_plus_bias_scaled = scaler.fit_transform(
housing_data_plus_bias)
X = housing_data_plus_bias_scaled
print(X.dtype)
y = housing.target.reshape(-1, 1)
n_epochs = 100
learning_rate = 0.01
batch_size = 100
n_batches = int(np.ceil(m / batch_size))
grad = 'optimizer'
X_batch = tf.compat.v1.placeholder(tf.float32,
shape=(None, n + 1),
name='X_batch')
y_batch = tf.compat.v1.placeholder(tf.float32,
shape=(None, 1),
name='y_batch')
theta = tf.Variable(tf.compat.v1.random_uniform([n + 1, 1], -1.0, 1.0),
name="theta")
y_pred = tf.matmul(X_batch, theta, name="predictions")
error = y_pred - y_batch
mse = tf.reduce_mean(tf.square(error), name="mse")
if grad == 'autodiff':
gradients = tf.gradients(mse, [theta])[0]
training_op = tf.compat.v1.assign(theta,
theta - learning_rate * gradients)
elif grad == 'optimizer':
optimizer = tf.compat.v1.train.GradientDescentOptimizer(
learning_rate=learning_rate)
training_op = optimizer.minimize(mse)
else:
gradients = 2 / m * tf.matmul(tf.transpose(X_batch), error)
training_op = tf.compat.v1.assign(theta,
theta - learning_rate * gradients)
init = tf.compat.v1.global_variables_initializer()
mse_summary = tf.compat.v1.summary.scalar('MSE', mse)
file_writer = tf.compat.v1.summary.FileWriter(
logdir, tf.compat.v1.get_default_graph())
# for saving the model
#saver = tf.compat.v1.train.Saver()
with tf.compat.v1.Session() as sess:
sess.run(init)
for epoch in range(n_epochs):
#if epoch % 100 == 0:
#print("Epoch: ", epoch, "MSE: ", mse.eval())
#save_path = saver.save(sess, 'model.ckpt')
for batch_index in range(n_batches):
X_batch_fetched, y_batch_fetched = fetch_batch(
X, y, batch_index, batch_size)
print('Epoch: {}, Batch: {}, MSE: {}'.format(
epoch, batch_index,
mse.eval(feed_dict={
X_batch: X_batch_fetched,
y_batch: y_batch_fetched
})))
if batch_index % 10 == 0:
summary_str = mse_summary.eval(feed_dict={
X_batch: X_batch_fetched,
y_batch: y_batch_fetched
})
step = epoch * n_batches + batch_index
file_writer.add_summary(summary_str, step)
sess.run(training_op,
feed_dict={
X_batch: X_batch_fetched,
y_batch: y_batch_fetched
})
best_theta = theta.eval()
from matplotlib import pyplot as plt
from matplotlib import cm
from sklearn.preprocessing import MinMaxScaler
from sklearn.preprocessing import minmax_scale
from sklearn.preprocessing import MaxAbsScaler
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import RobustScaler
from sklearn.preprocessing import Normalizer
from sklearn.preprocessing.data import QuantileTransformer
from sklearn.datasets import fetch_california_housing
print(__doc__)
dataset = fetch_california_housing()
X_full, y_full = dataset.data, dataset.target
# Take only 2 features to make visualization easier
# Feature of 0 has a long tail distribution.
# Feature 5 has a few but very large outliers.
X = X_full[:, [0, 5]]
distributions = [
('Unscaled data', X),
('Data after standard scaling',
StandardScaler().fit_transform(X)),
('Data after min-max scaling',
MinMaxScaler().fit_transform(X)),
('Data after max-abs scaling',
def fetch(*args, **kwargs):
return fetch_california_housing(*args, download_if_missing=False, **kwargs)
def get_dataset(dataset_name='breast_cancer'):
"""Retrieve one of the standard datasets in sklearn.
:param dataset_name: the dataset name to use from sklearn. Valid values are
`breast_cancer`, `digits`, `iris`, `wine` for classification and
`boston`, `diabetes` and `california` for regression. Default is `breast_cancer`.
:type dataset_name: str
:return: Five variables are returned. First is the dataset itself without
the target values; second includes the target values; third has
all the categorical columns; fourth has all the integer columns and
the last informs if it is a classification problem (True) or a regression
problem (False).
:rtype: np.array, np.array, np.array, np.array, Boolean
"""
if dataset_name == 'breast_cancer':
# loading the dataset
X, y = load_breast_cancer(return_X_y=True)
# informing categorical columns and their available values
categorical_columns = {}
integer_columns = []
is_classification = True
elif dataset_name == 'digits':
# loading the dataset
X, y = load_digits(return_X_y=True)
# informing categorical columns and their available values
categorical_columns = {}
integer_columns = list(range(64))
is_classification = True
elif dataset_name == 'iris':
# loading the dataset
X, y = load_iris(return_X_y=True)
# informing categorical columns and their available values
categorical_columns = {}
integer_columns = []
is_classification = True
elif dataset_name == 'wine':
# loading the dataset
X, y = load_wine(return_X_y=True)
# informing categorical columns and their available values
categorical_columns = {}
integer_columns = [4, 12]
is_classification = True
elif dataset_name == 'boston':
X, y = load_boston(return_X_y=True)
# informing categorical columns and their available values
categorical_columns = {3: [0, 1]}
integer_columns = [8, 9]
is_classification = False
elif dataset_name == 'diabetes':
# loading the dataset
X, y = load_diabetes(return_X_y=True)
# informing categorical columns and their available values
categorical_columns = {1: [0.05068012, -0.04464164]}
integer_columns = []
is_classification = False
elif dataset_name == 'california':
# loading the dataset
X, y = fetch_california_housing(return_X_y=True)
# informing categorical columns and their available values
categorical_columns = {}
integer_columns = [1, 4]
is_classification = False
return X, y, categorical_columns, integer_columns, is_classification
from matplotlib import pyplot as plt
from matplotlib import cm
from sklearn.preprocessing import MinMaxScaler
from sklearn.preprocessing import minmax_scale
from sklearn.preprocessing import MaxAbsScaler
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import RobustScaler
from sklearn.preprocessing import Normalizer
from sklearn.preprocessing.data import QuantileTransformer
from sklearn.datasets import fetch_california_housing
print(__doc__)
dataset = fetch_california_housing()
X_full, y_full = dataset.data, dataset.target
# Take only 2 features to make visualization easier
# Feature of 0 has a long tail distribution.
# Feature 5 has a few but very large outliers.
X = X_full[:, [0, 5]]
distributions = [
('Unscaled data', X),
('Data after standard scaling',
StandardScaler().fit_transform(X)),
('Data after min-max scaling',
MinMaxScaler().fit_transform(X)),
('Data after max-abs scaling',
#!/usr/bin/env python
# -*- coding: utf-8 -*-
from sklearn.datasets import fetch_california_housing
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import RandomizedSearchCV
from sklearn.model_selection import train_test_split
from scipy.stats import randint
X, y = fetch_california_housing(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)
# define the parameter space that will be searched over
param_distributions = {
'n_estimators': randint(1, 5),
'max_depth': randint(5, 10)
}
# now create a searchCV object and fit it to the data
search = RandomizedSearchCV(estimator=RandomForestRegressor(random_state=0),
n_iter=5,
param_distributions=param_distributions,
random_state=0)
search.fit(X_train, y_train)
print(search.best_params_)
# the search object now acts like a normal random forest estimator
# with max_depth=9 and n_estimators=4
print(search.score(X_test, y_test))
First, we'll load some packages.
'''
# -
import pandas as pd
import xgboost
import shap
from sklearn import datasets as ds
# + [markdown]
'''
Next, we'll load some data. Let's use the California housing data set.
'''
# +
calif_house_data = ds.fetch_california_housing()
print('\n')
print(calif_house_data['data'])
print('\n')
print(calif_house_data['target'])
print('\n')
print(calif_house_data['feature_names'])
print('\n')
print(calif_house_data['DESCR'])
print('\n')
print('Features - # rows, # columns:', calif_house_data['data'].shape)
print('\n')
print('Target variable - # rows:', calif_house_data['target'].shape)
# + [markdown]
from sklearn.datasets import fetch_california_housing
california = fetch_california_housing() # Bunch object
'''
print(california.DESCR)
print(california.data.shape)
print(california.target.shape)
print(california.feature_names)
'''
import pandas as pd
pd.set_option("precision", 4) # 4 digitit precision for floats.
pd.set_option("max_columns", 9) # display up to 9 columns in DataFrame outputs
pd.set_option("display.width", None) #auto-detect the display width for wrapping
#creates the initial DataFrame using the data in california.data and with the
# column names specified based on the features of the sample
california_df = pd.DataFrame(california.data, columns=california.feature_names)
# add a column to the dataframe for the mdian house values stored in california.target
california_df["MedHouseValue"] = pd.Series(california.target)
print(california_df.head())
# using the describe method of dataframes we can get some statistical information
def load_dataset(name,
num_features=5,
random_state=42,
flatten=False,
show_corr_matrix=True,
show_subplots=True):
'''
Args:
name (str): name of dataset ('mnist', 'fmnist', 'iris', 'breast_cancer', 'diabetes', 'housing')
num_features (int): number of features to view in subplots, if None then num_features includes all features
random_state (int): specify random state
flatten (bool): returns image with shape (-1, 28, 28) if True, else returns image with shape (-1, 784)
show_corr_matrix (bool): whether to show correlation matrix
show_subplots (bool): whether to show subplots comparing the num_features
Returns:
(X_train, y_train): training data (numpy arrays)
(X_test, y_test): testing data (20% of total data)
col_names (list): feature names
'''
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
if name == 'mnist':
(X_train, y_train), (X_test, y_test) = mnist.load_data()
if flatten:
X_train = X_train.reshape(-1, 784)
X_test = X_test.reshape(-1, 784)
X_train = X_train / 255.0
X_test = X_test / 255.0
return (X_train, y_train), (X_test, y_test)
elif name == 'fashion_mnist' or name == 'fmnist':
(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()
if flatten:
X_train = X_train.reshape(-1, 784)
X_test = X_test.reshape(-1, 784)
return (X_train, y_train), (X_test, y_test)
elif name == 'iris':
data = datasets.load_iris()
X = data.data
y = data.target
col_names = data.feature_names
df = pd.DataFrame(X, columns=col_names)
num = num_features
if num_features == None:
num = len(col_names)
if show_subplots:
plot_subplots(df, col_names, num)
if show_corr_matrix:
corr_img = correlations(df, col_names)
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.2,
random_state=42)
elif name == 'cancer' or name == 'breast_cancer':
data = datasets.load_breast_cancer()
X = data.data
y = data.target
col_names = data.feature_names
df = pd.DataFrame(X, columns=col_names)
num = num_features
if num_features == None:
num = len(col_names)
if show_subplots:
plot_subplots(df, col_names, num)
if show_corr_matrix:
corr_img = correlations(df, col_names)
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.2,
random_state=42)
elif name == 'diabetes':
data = datasets.load_diabetes()
X = data.data
y = data.target
col_names = data.feature_names
df = pd.DataFrame(X, columns=col_names)
num = num_features
if num_features == None:
num = len(col_names)
if show_subplots:
plot_subplots(df, col_names, num)
if show_corr_matrix:
corr_img = correlations(df, col_names)
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.2,
random_state=42)
elif name == 'house' or name == 'california_housing' or name == 'housing':
data = datasets.fetch_california_housing()
X = data.data
y = data.target
col_names = data.feature_names
df = pd.DataFrame(X, columns=col_names)
num = num_features
if num_features == None:
num = len(col_names)
if show_subplots:
plot_subplots(df, col_names, num)
if show_corr_matrix:
corr_img = correlations(df, col_names)
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.2,
random_state=42)
if show_corr_matrix:
plt.title("Correlation Matrix")
return (X_train, y_train), (X_test, y_test), col_names
# regression MLP model
import tensorflow as tf
from sklearn import model_selection, preprocessing
from sklearn import datasets
# Using the sklearn california housing data
housing = datasets.fetch_california_housing()
X_train_full, X_test, y_train_full, y_test = model_selection.train_test_split(
housing.data, housing.target)
X_train, X_valid, y_train, y_valid = model_selection.train_test_split(
X_train_full, y_train_full)
# Standard scaling the data
scaler = preprocessing.StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_valid_scaled = scaler.transform(X_valid)
X_test_scaled = scaler.transform(X_test)
# Creating a simple model
model = tf.keras.models.Sequential([
tf.keras.layers.Dense(30, activation="selu",
input_shape=X_train.shape[1:]),
tf.keras.layers.Dense(10, activation="selu"),
tf.keras.layers.Dense(1)
])
# Compiling using huber loss
model.compile(loss=tf.keras.losses.Huber(), optimizer="adam")
# training
history = model.fit(X_train,
metrics = metric(torch.clamp(y_hat, min=0),
torch.clamp(y[:, 1], min=0), **metric_params)
else:
metrics = metric(y_hat, y[:, 1], **metric_params)
self.log(
f"{tag}_{metric_str}",
metrics,
on_epoch=True,
on_step=False,
logger=True,
prog_bar=True,
)
return metrics
dataset = fetch_california_housing(data_home="data", as_frame=True)
dataset.frame["HouseAgeBin"] = pd.qcut(dataset.frame["HouseAge"], q=4)
dataset.frame.HouseAgeBin = "age_" + dataset.frame.HouseAgeBin.cat.codes.astype(
str)
test_idx = dataset.frame.sample(int(0.2 * len(dataset.frame)),
random_state=42).index
test = dataset.frame[dataset.frame.index.isin(test_idx)]
train = dataset.frame[~dataset.frame.index.isin(test_idx)]
epochs = 15
batch_size = 128
steps_per_epoch = int((len(train) // batch_size) * 0.9)
data_config = DataConfig(
target=["HouseAgeBin"] + dataset.target_names,
continuous_cols=[
import pylab as pl
import numpy as np
from matplotlib import pyplot as plt
from sompy.sompy import SOMFactory
from sklearn.datasets import fetch_california_housing
from sompy.visualization.mapview import View2D
from sompy.visualization.bmuhits import BmuHitsView
data = fetch_california_housing()
descr = data.DESCR
names = fetch_california_housing().feature_names+["HouseValue"]
data = np.column_stack([data.data, data.target])
print(descr)
print( "FEATURES: ", ", ".join(names))
sm = SOMFactory().build(data, normalization = 'var', initialization='random', component_names=names)
sm.train(n_job=1, verbose=False, train_rough_len=2, train_finetune_len=5)
topographic_error = sm.calculate_topographic_error()
quantization_error = np.mean(sm._bmu[1])
print ("Topographic error = %s; Quantization error = %s" % (topographic_error, quantization_error))
view2D = View2D(10, 10, 'rand data', text_size=10)
view2D.show(sm, col_sz=4, which_dim='all', denormalize=True)
vhts = BmuHitsView(10, 10, 'Hits Map', text_size=7)
def test_run():
X, y = fetch_california_housing(data_home=TEST_FOLDER, return_X_y=True)
data = pd.DataFrame(X)
data['target'] = y
Pipeline = compose([('scaler', StandardScaler),
('lin_reg', LinearRegression)])
search_spaces = [
SearchSpace(id='Linear Regression', model_class=LinearRegression),
SearchSpace(id='Lasso', model_class=Lasso),
SearchSpace(id='Pipeline',
model_class=Pipeline,
parameters_values=dict(
scaler__with_mean=[True, False],
scaler__with_std=[True, False],
lin_reg__fit_intercept=[True, False],
lin_reg__normalize=[True, False]))
]
config = Config(local_dir=TEST_FOLDER,
problem_type='regression',
score_function=r2_score,
search_spaces=search_spaces,
ensemble_id='Ensemble',
stagnation=1)
engine = Engine(config)
train_data, test_data = train_test_split(data, test_size=0.2)
train_data_original, test_data_original = train_data.copy(
), test_data.copy()
engine.load_train_data(train_data, 'target')
engine.load_test_data(test_data)
if engine.is_running():
raise AssertionError()
engine.restart()
sleep(2)
if not engine.is_running():
raise AssertionError()
sleep(5)
status = engine.request_status()
if len(status.scores) != len(search_spaces) + 1 or \
len(status.ensemble_weights) != len(search_spaces):
raise AssertionError()
if status.train_predictions.shape[0] != train_data.shape[0]:
raise AssertionError()
if status.test_predictions.shape[0] != test_data.shape[0]:
raise AssertionError()
for base_model in status.base_models.values():
for feature in base_model['features']:
if feature not in test_data.columns or feature not in train_data.columns:
raise AssertionError()
engine.interrupt()
if engine.is_running():
raise AssertionError()
engine.clean_test_data(restart=True)
sleep(5)
if not engine.is_running():
raise AssertionError()
engine.shuffle_train_data(restart=True)
sleep(5)
status = engine.request_status()
if status.test_predictions is not None:
raise AssertionError()
engine.interrupt()
status.build_report()
status.build_report(include_features=True)
pd.testing.assert_frame_equal(train_data, train_data_original)
pd.testing.assert_frame_equal(test_data, test_data_original)
# models result in more powerful and robust models with less hassle.
#
# We will start by loading the california housing dataset. We recall that the
# goal in this dataset is to predict the median house value in some district
# in California based on demographic and geographic data.
# %% [markdown]
# ```{note}
# If you want a deeper overview regarding this dataset, you can refer to the
# Appendix - Datasets description section at the end of this MOOC.
# ```
# %%
from sklearn.datasets import fetch_california_housing
data, target = fetch_california_housing(as_frame=True, return_X_y=True)
target *= 100 # rescale the target in k$
# %% [markdown]
# ```{caution}
# Here and later, we use the name `data` and `target` to be explicit. In
# scikit-learn documentation, `data` is commonly named `X` and `target` is
# commonly called `y`.
# %% [markdown]
# We will check the statistical performance of decision tree regressor with
# default parameters.
# %%
from sklearn.model_selection import cross_validate
from sklearn.tree import DecisionTreeRegressor
import tensorflow as tf
import numpy as np
from sklearn.datasets import fetch_california_housing
from sklearn.preprocessing import StandardScaler
n_epochs = 100000
learning_rate = 0.001
housing = fetch_california_housing(
data_home="C:/Users/28542/scikit_learn_data", download_if_missing=True)
m, n = housing.data.shape
print(m, n)
housing_data_plus_bias = np.c_[np.ones((m, 1)), housing.data]
# 可以使用TensorFlow或者Numpy或者sklearn的StandardScaler去进行归一化
# StandardScaler默认就做了方差归一化，和均值归一化，这两个归一化的目的都是为了更快的进行梯度下降
# 你如何构建你的训练集，你训练除了的模型，就具备什么样的功能！
scaler = StandardScaler().fit(housing_data_plus_bias)
scaled_housing_data_plus_bias = scaler.transform(housing_data_plus_bias)
X = tf.constant(scaled_housing_data_plus_bias, dtype=tf.float32, name='X')
y = tf.constant(housing.target.reshape(-1, 1), dtype=tf.float32, name='y')
# random_uniform函数创建图里一个节点包含随机数值，给定它的形状和取值范围，就像numpy里面rand()函数
theta = tf.Variable(tf.random_uniform([n + 1, 1], -1.0, 1.0), name='theta')
y_pred = tf.matmul(X, theta, name="predictions")
error = y_pred - y
mse = tf.reduce_mean(tf.square(error), name="mse")
# 梯度的公式：(y_pred - y) * xj
gradients = 2 / m * tf.matmul(tf.transpose(X), error)
# 赋值函数对于BGD来说就是 theta_new = theta - (learning_rate * gradients)
training_op = tf.assign(theta, theta - learning_rate * gradients)
from sklearn.datasets import fetch_california_housing
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from tensorflow import keras
housing = fetch_california_housing()
X_train_full, X_test, y_train_full, y_test = train_test_split(housing.data,
housing.target,
random_state=0)
X_train, X_valid, y_train, y_valid = train_test_split(X_train_full,
y_train_full)
scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_valid = scaler.transform(X_valid)
X_test = scaler.transform(X_test)
# bacis version
# input = keras.layers.Input(shape=X_train.shape[1:])
# hidden1 = keras.layers.Dense(30, activation="relu")(input)
# hidden2 = keras.layers.Dense(30, activation="relu")(hidden1)
# concat = keras.layers.Concatenate()[input, hidden2]
# output = keras.layers.Dense(1)(concat)
# model = keras.models.Model(input=[input], output=[output])
# send a subset of the features throuph the wide path, and a different subset throuph the deep path
input_A = keras.layers.Input(shape=[5])
input_B = keras.layers.Input(shape=[6])
hidden1 = keras.layers.Dense(30, activation="relu")(input_B)
def test_california_housing_oob():
X, y = fetch_california_housing(return_X_y=True)
run_regression_test(X, y, min_training_score=.79, grace=0.15, oob=True)
import tensorflow as tf
import numpy as np
from sklearn.datasets import fetch_california_housing
# 立刻下载数据集
housing = fetch_california_housing(data_home="C:/Users/28542/scikit_learn_data", download_if_missing=False)
# 获得X数据行数和列数
m, n = housing.data.shape
# 这里添加一个额外的bias输入特征(x0=1)到所有的训练数据上面，因为使用的numpy所有会立即执行
housing_data_plus_bias = np.c_[np.ones((m, 1)), housing.data]
# 创建两个TensorFlow常量节点X和y，去持有数据和标签
X = tf.constant(housing_data_plus_bias, dtype=tf.float32, name='X')
y = tf.constant(housing.target.reshape(-1, 1), dtype=tf.float32, name='y')
# 使用一些TensorFlow框架提供的矩阵操作去求theta
XT = tf.transpose(X)
# 解析解一步计算出最优解
theta = tf.matmul(tf.matmul(tf.matrix_inverse(tf.matmul(XT, X)), XT), y)
with tf.Session() as sess:
theta_value = theta.eval() # sess.run(theta)
print(theta_value)
def test_california_housing():
X, y = fetch_california_housing(return_X_y=True)
run_regression_test(X, y, ntrials=10, grace=0.19)
from sklearn.datasets import fetch_california_housing from sklearn.preprocessing import StandardScaler # TensorFlow为我们去计算梯度，但是同时也给了我们更方便的求解方式 # 它提供给我们与众不同的，有创意的一些优化器，包括梯度下降优化器 # 替换前面代码相应的行，并且一切工作正常 # 设定超参数，Grid Search进行栅格搜索，其实说白了就是排列组合找到Loss Function最小的时刻 # 的那组超参数结果 n_epochs = 1000 learning_rate = 0.01 # 读取数据，这里读取数据是一下子就把所有数据交给X，Y节点，所以下面去做梯度下降的时候 # BGD = Batch Gradient Decrease ，如果面向数据集比较大的时候，我们倾向与 Mini GD housing = fetch_california_housing() m, n = housing.data.shape housing_data_plus_bias = np.c_[np.ones((m, 1)), housing.data] # 可以使用TensorFlow或者Numpy或者sklearn的StandardScaler去进行归一化 scaler = StandardScaler().fit(housing_data_plus_bias) scaled_housing_data_plus_bias = scaler.transform(housing_data_plus_bias) # 下面部分X，Y最后用placeholder可以改成使用Mini BGD # 构建计算的图 X = tf.constant(scaled_housing_data_plus_bias, dtype=tf.float32, name='X') y = tf.constant(housing.target.reshape(-1, 1), dtype=tf.float32, name='y') # random_uniform函数创建图里一个节点包含随机数值，给定它的形状和取值范围，就像numpy里面rand()函数 theta = tf.Variable(tf.random_uniform([n + 1, 1], -1.0, 1.0), name='theta') y_pred = tf.matmul(X, theta, name="predictions") error = y_pred - y
from sklearn.experimental import enable_iterative_imputer # noqa
from sklearn.datasets import fetch_california_housing
from sklearn.impute import SimpleImputer
from sklearn.impute import IterativeImputer
from sklearn.linear_model import BayesianRidge
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import ExtraTreesRegressor
from sklearn.neighbors import KNeighborsRegressor
from sklearn.pipeline import make_pipeline
from sklearn.model_selection import cross_val_score
N_SPLITS = 5
rng = np.random.RandomState(0)
X_full, y_full = fetch_california_housing(return_X_y=True)
# ~2k samples is enough for the purpose of the example.
# Remove the following two lines for a slower run with different error bars.
X_full = X_full[::10]
y_full = y_full[::10]
n_samples, n_features = X_full.shape
# Estimate the score on the entire dataset, with no missing values
br_estimator = BayesianRidge()
score_full_data = pd.DataFrame(
cross_val_score(
br_estimator, X_full, y_full, scoring="neg_mean_squared_error", cv=N_SPLITS
),
columns=["Full Data"],
)
parser.add_argument('--test_size', type=float, default=0.1, help="TestSize")
args = parser.parse_args()
print("\n" + "Arguments are: " + "\n")
print(args)
import time
from tqdm import tqdm
# Loading datasets
if args.ds == 'boston':
X, y = load_boston()['data'], load_boston()['target']
elif args.ds == 'diabetes':
X, y = load_diabetes()['data'], load_diabetes()['target']
elif args.ds == 'cali':
X, y = fetch_california_housing()['data'], fetch_california_housing()['target']
else:
X, y = make_regression(n_samples=args.n_samples, n_features=args.n_feats, noise=args.noise, random_state=0)
start = time.time()
array_wob, array_wb = experiment_1(X=X, y=y, n_train_iter=args.n_train_iter, n_average=args.n_average,
num_N=args.N, test_size=args.test_size)
end = time.time()
print(str(end - start) + ' seconds')
print('finish')
with open(f'results_mse_{args.ds}|{args.noise}|{args.n_average}|{args.n_feats}|{args.n_samples}_'
f'mlp_test_size={args.test_size}.txt', 'w') as f:
f.write(str(0) + '|' + str(np.mean(array_wob)) + '\n')
for i, el in zip(range(1, len(array_wb)), array_wb):
f.write(str(i) + '|' + str(np.mean(el)) + '\n')
f.close()
def fetch(*args, **kwargs):
return fetch_california_housing(*args, download_if_missing=False, **kwargs)
