housing.info()
# 分析：数据集中共有 20640 个实例，按照机器学习的标准这个数据量很小，但是非常适合入门。
# 我们注意到总卧室数只有 20433 个非空值，这意味着有 207 个街区缺少这个值。我们将在后面对它进行处理。

housing["ocean_proximity"].value_counts()

housing.plot(kind="scatter", x="longitude", y="latitude", alpha=0.4, s=housing["population"]/100, label="population",c="median_house_value", cmap=plt.get_cmap("jet"), colorbar=True,)
# 每个圈的半径表示街区的人口（选项`s`），颜色代表价格（选项`c`）
plt.legend()

corr_matrix = housing.corr()
corr_matrix["median_house_value"].sort_values(ascending=False)

attributes = ["median_house_value", "median_income", "total_rooms","housing_median_age"]
pd.plotting.scatter_matrix(housing[attributes], figsize=(12, 8))

某些属性本身没有用，与其它属性结合起来就有用了。比如下面房间数与房主数本身没有用，相除得出的每户的房间数更有用。所以实际操作中需要各种属性组合，然后比较相关系数。

housing["rooms_per_household"] = housing["total_rooms"]/housing["households"]
housing["bedrooms_per_room"] = housing["total_bedrooms"]/housing["total_rooms"]
housing["population_per_household"]=housing["population"]/housing["households"]corr_matrix = housing.corr()
corr_matrix["median_house_value"].sort_values(ascending=False)

为机器学习算法准备数据

不要手工来做，需要一些函数。原因：

• 函数可以在任何数据集上方便重复地数据转换
• 慢慢建立一个函数库，在未来的项目中重复使用
• 可以方便尝试多种数据转换

第一步，将特征和标签分开

housing = strat_train_set.drop("median_house_value", axis=1)
housing_labels = strat_train_set["median_house_value"]

数据清洗

缺失值（本例使用total_bedrooms）：

• 去掉对应的街区 dropna()
• 去掉整个属性 drop()
• 进行赋值（0，平均值，中位数等）fillna() 使用该方法时记得保存平均值或者中位值等，后面测试集也要填充

housing.dropna(subset=["total_bedrooms"])  # 选项1
housing.drop("total_bedrooms", axis=1) # 选项2
median = housing["total_bedrooms"].median()
housing["total_bedrooms"].fillna(median) # 选项3

from sklearn.preprocessing import Imputerimputer = Imputer(strategy="median")# 因为只有数值属性才能算出中位数，我们需要创建一份不包括文本属性`ocean_proximity`的数据副本
housing_num = housing.drop("ocean_proximity", axis=1)# 用`fit()`方法将`imputer`实例拟合到训练数据
imputer.fit(housing_num)

X = imputer.transform(housing_num)
housing_tr = pd.DataFrame(X, columns=housing_num.columns)

from sklearn.preprocessing import LabelEncoderencoder = LabelEncoder()
housing_cat = housing["ocean_proximity"]
housing_cat_encoded = encoder.fit_transform(housing_cat)
housing_cat_encoded

原理是创建一个二元属性，当分类是<1H OCEAN，该属性为 1（否则为 0），当分类是INLAND，另一个属性等于 1（否则为 0），以此类推。这称作独热编码（One-Hot Encoding），因为只有一个属性会等于 1（热），其余会是 0（冷）

注意：fit_transform()用于 2D 数组，而housing_cat_encoded`是一个 1D 数组，所以需要将其变形

from sklearn.preprocessing import OneHotEncoderencoder = OneHotEncoder()
housing_cat_1hot = encoder.fit_transform(housing_cat_encoded.reshape(-1,1))
# housing_cat_1hot结果是一个稀疏矩阵，即只存储非零项。这是为了当分类很多时节省内存。 将其转换成numpy数组需要用到toarray函数
housing_cat_1hot.toarray()

from sklearn.preprocessing import LabelBinarizerencoder = LabelBinarizer()
housing_cat_1hot = encoder.fit_transform(housing_cat)
housing_cat_1hot

from sklearn.base import BaseEstimator, TransformerMixin
rooms_ix, bedrooms_ix, population_ix, household_ix = 3, 4, 5, 6class CombinedAttributesAdder(BaseEstimator, TransformerMixin):def __init__(self, add_bedrooms_per_room = True): # no *args or **kargsself.add_bedrooms_per_room = add_bedrooms_per_roomdef fit(self, X, y=None):return self # nothing else to dodef transform(self, X, y=None):rooms_per_household = X[:, rooms_ix] / X[:, household_ix]population_per_household = X[:, population_ix] / X[:, household_ix]if self.add_bedrooms_per_room:bedrooms_per_room = X[:, bedrooms_ix] / X[:, rooms_ix]return np.c_[X, rooms_per_household, population_per_household,bedrooms_per_room]else:return np.c_[X, rooms_per_household, population_per_household]attr_adder = CombinedAttributesAdder(add_bedrooms_per_room=False)
housing_extra_attribs = attr_adder.transform(housing.values)

from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScalernum_pipeline = Pipeline([('imputer', Imputer(strategy="median")),('attribs_adder', CombinedAttributesAdder()),('std_scaler', StandardScaler()),])
housing_num_tr = num_pipeline.fit_transform(housing_num)

from sklearn.base import BaseEstimator, TransformerMixinclass DataFrameSelector(BaseEstimator, TransformerMixin):def __init__(self, attribute_names):self.attribute_names = attribute_namesdef fit(self, X, y=None):return selfdef transform(self, X):return X[self.attribute_names].values

# 由于sklearn更新使得LabelBinarizer的fit_transform只能接受两个参数，直接运行的话会报错，所以重写一个转化器，增加一个参数。
from sklearn.base import TransformerMixin #gives fit_transform method for free
class MyLabelBinarizer(TransformerMixin):def __init__(self, *args, **kwargs):self.encoder = LabelBinarizer(*args, **kwargs)def fit(self, x, y=0):self.encoder.fit(x)return selfdef transform(self, x, y=0):return self.encoder.transform(x)

from sklearn.pipeline import FeatureUnionnum_attribs = list(housing_num)
cat_attribs = ["ocean_proximity"]num_pipeline = Pipeline([('selector', DataFrameSelector(num_attribs)),('imputer', Imputer(strategy="median")),('attribs_adder', CombinedAttributesAdder()),('std_scaler', StandardScaler()),])cat_pipeline = Pipeline([('selector', DataFrameSelector(cat_attribs)),('Mylabel_binarizer', MyLabelBinarizer()),])full_pipeline = FeatureUnion(transformer_list=[("num_pipeline", num_pipeline),("cat_pipeline", cat_pipeline),])

housing_prepared = full_pipeline.fit_transform(housing)
housing_prepared[0]

from sklearn.linear_model import LinearRegressionlin_reg = LinearRegression()
lin_reg.fit(housing_prepared, housing_labels)

#结果0.6558010255907188

from sklearn.metrics import mean_squared_error housing_predictions = lin_reg.predict(housing_prepared)
lin_mse = mean_squared_error(housing_labels, housing_predictions)
lin_rmse = np.sqrt(lin_mse)
lin_rmse

from sklearn.model_selection import cross_val_scorelin_scores = cross_val_score(lin_reg, housing_prepared, housing_labels,scoring="neg_mean_squared_error", cv=10)
lin_rmse_scores = np.sqrt(-lin_scores)# lin_scores越大越好，取负值开方后的lin_rmse_scores越小越好

def display_scores(scores):print("Scores:", scores)print("Mean:", scores.mean())print("Standard deviation:", scores.std())

display_scores(lin_rmse_scores)
# Mean就是RMSE

from sklearn.tree import DecisionTreeRegressortree_reg = DecisionTreeRegressor()
tree_reg.fit(housing_prepared, housing_labels)

housing_predictions = tree_reg.predict(housing_prepared)
tree_mse = mean_squared_error(housing_labels, housing_predictions)
tree_rmse = np.sqrt(tree_mse)
tree_rmse#结果 0.0

scores = cross_val_score(tree_reg, housing_prepared, housing_labels, scoring="neg_mean_squared_error", cv=10)
tree_rmse_scores = np.sqrt(-scores)display_scores(tree_rmse_scores)

from sklearn.ensemble import RandomForestRegressorforest_reg = RandomForestRegressor()
forest_reg.fit(housing_prepared, housing_labels)

housing_predictions = forest_reg.predict(housing_prepared)
forest_mse = mean_squared_error(housing_labels, housing_predictions)
forest_rmse = np.sqrt(forest_mse)
forest_rmse#22088.901578494966

scores = cross_val_score(forest_reg, housing_prepared, housing_labels, scoring="neg_mean_squared_error", cv=10)
forest_rmse_scores = np.sqrt(-scores)display_scores(forest_rmse_scores)

from sklearn.model_selection import GridSearchCVparam_grid = [{'n_estimators': [3, 10, 30], 'max_features': [2, 4, 6, 8]},{'bootstrap': [False], 'n_estimators': [3, 10], 'max_features': [2, 3, 4]},]forest_reg = RandomForestRegressor()grid_search = GridSearchCV(forest_reg, param_grid, cv=5, scoring='neg_mean_squared_error')
grid_search.fit(housing_prepared, housing_labels)

grid_search.best_params_

grid_search.best_estimator_

# 得到评估得分
cvres = grid_search.cv_results_
for mean_score, params in zip(cvres["mean_test_score"], cvres["params"]):print(np.sqrt(-mean_score), params)

可以指出每个属性对于做出准确预测的相对重要性

feature_importances = grid_search.best_estimator_.feature_importances_
feature_importances

extra_attribs = ["rooms_per_hhold", "pop_per_hhold", "bedrooms_per_room"]cat_one_hot_attribs = list(encoder.classes_)
attributes = num_attribs + extra_attribs + cat_one_hot_attribssorted(zip(feature_importances,attributes), reverse=True)

由上可以看出，字段中ISLAND最重要，其它的几个可以删掉

用测试集评估系统

final_model = grid_search.best_estimator_X_test = strat_test_set.drop("median_house_value", axis=1)
y_test = strat_test_set["median_house_value"].copy()
X_test_prepared = full_pipeline.transform(X_test)final_predictions = final_model.predict(X_test_prepared)final_mse = mean_squared_error(y_test, final_predictions)
final_rmse = np.sqrt(final_mse)display_scores(final_rmse)