机器学习系列------超参数（完成本次机器学习任务）

alltolove

        还是继续复制过来以前的代码：

import pandas as pd
import os
import numpy as np
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import Imputer
from sklearn.base import BaseEstimator,TransformerMixin
from sklearn.preprocessing import LabelBinarizer
from sklearn.pipeline import FeatureUnion
HOUSING_PATH="datasets/housing"
def load_housing_data(housing_path=HOUSING_PATH):
    csv_path=os.path.join(housing_path,"housing.csv")
    return pd.read_csv(csv_path)
housing=load_housing_data()



housing["income_cat"]=np.ceil(housing["median_income"]/1.5)
housing["income_cat"].where(housing["income_cat"]<5,5.0,inplace=True)



from sklearn.model_selection import StratifiedShuffleSplit
split=StratifiedShuffleSplit(n_splits=1,test_size=0.2,random_state=42)
for train_index,test_index in split.split(housing,housing["income_cat"]):
    strat_train_set=housing.loc[train_index]
    strat_test_set=housing.loc[test_index]



for set in (strat_train_set,strat_test_set):
    set.drop(["income_cat"],axis=1,inplace=True)
    
housing=strat_train_set.drop("median_house_value",axis=1)
housing_labels=strat_train_set["median_house_value"].copy()
housing_num=housing.drop("ocean_proximity",axis=1)

rooms_ix,bedroom_ix,population_ix,household_ix=3,4,5,6
class CombinedAttributesAdder(BaseEstimator,TransformerMixin):
    def __init__(self,add_bedrooms_per_room=True):
        self.add_bedrooms_per_room=add_bedrooms_per_room
    def fit(self,X,y=None):
        return self
    def 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[:,bedroom_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]                 
        
class MyLabelBinarizer(TransformerMixin):
    def __init__(self, *args, **kwargs):
        self.encoder = LabelBinarizer(*args, **kwargs)
    def fit(self, x, y=0):
        self.encoder.fit(x)
        return self
    def transform(self, x, y=0):
        return self.encoder.transform(x)
    
class DataFrameSelector(BaseEstimator,TransformerMixin):
    def __init__(self,attribute_names):
        self.attribute_names=attribute_names
    def fit(self,X,y=None):
        return self
    def transform(self,X):
        return X[self.attribute_names].values

num_attribs=list(housing_num)
cat_attribs=["ocean_proximity"]
num_pipline=Pipeline([
        ('selector',DataFrameSelector(num_attribs)),
        ('imputer',Imputer(strategy="median")),
        ('attribus_adder',CombinedAttributesAdder()),
        ('std_scaler',StandardScaler()),
    ])
cat_pipline=Pipeline([
        ('selector',DataFrameSelector(cat_attribs)),
        ('label_binarizer',MyLabelBinarizer()),
    ])
full_pipline=FeatureUnion(transformer_list=[
        ("num_pipline",num_pipline),
        ("cat_pipline",cat_pipline),
    ])
housing_prepared=full_pipline.fit_transform(housing)

        然后为我们的模型设置超参数（就是随机森林里树的个数，训练集有几组特征值等等一系列的参数）：

from sklearn.model_selection import GridSearchCV
from sklearn.ensemble import RandomForestRegressor

param_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)

        然后看看以上方法为我们预测出来的超参数是什么：

grid_search.best_params_

        写完上面这行执行后会出现如下数据：
{'max_features': 8, 'n_estimators': 30}
        这就是我们要用的最优参数，最大特征值是8组，决策树的个数是30个，会使我们预测的数据更准确。继续看下对比后的结果：

cvres=grid_search.cv_results_
for mean_score,params in zip(cvres["mean_test_score"],cvres["params"]):
    print(np.sqrt(-mean_score),params)

        执行后会显示以下结果：
64539.9065965 {'n_estimators': 3, 'max_features': 2}
55401.3638545 {'n_estimators': 10, 'max_features': 2}
52849.5504873 {'n_estimators': 30, 'max_features': 2}
60488.5008218 {'n_estimators': 3, 'max_features': 4}
53228.6746741 {'n_estimators': 10, 'max_features': 4}
50746.7001414 {'n_estimators': 30, 'max_features': 4}
59350.5028588 {'n_estimators': 3, 'max_features': 6}
52568.7835547 {'n_estimators': 10, 'max_features': 6}
50013.8405865 {'n_estimators': 30, 'max_features': 6}
59451.5995633 {'n_estimators': 3, 'max_features': 8}
52092.3880668 {'n_estimators': 10, 'max_features': 8}
49857.1401575 {'n_estimators': 30, 'max_features': 8}
61897.0944086 {'n_estimators': 3, 'bootstrap': False, 'max_features': 2}
54548.6814397 {'n_estimators': 10, 'bootstrap': False, 'max_features': 2}
60637.0802455 {'n_estimators': 3, 'bootstrap': False, 'max_features': 3}
53100.3544235 {'n_estimators': 10, 'bootstrap': False, 'max_features': 3}
58487.3207429 {'n_estimators': 3, 'bootstrap': False, 'max_features': 4}
51741.4498429 {'n_estimators': 10, 'bootstrap': False, 'max_features': 4}
        果然最优的就是刚才的那个，第一个均差最大的是最差的。然后看看特征的重要程度：

feature_importances=grid_search.best_estimator_.feature_importances_
feature_importances

        这个输出结果不是太容易看，只是一组数据这里就不复制了，然后我们用更直观的方法看看：

from sklearn.preprocessing import LabelEncoder
encoder=LabelEncoder()
housing_cat=housing["ocean_proximity"]
housing_cat_encoded=encoder.fit_transform(housing_cat)
extra_attribs=["rooms_per_household","pop_per_household","bedrooms_per_room"]
cat_one_hot_attribs=list(encoder.classes_)
attribus=num_attribs+extra_attribs+cat_one_hot_attribs
sorted(zip(feature_importances,attribus),reverse=True)

        输出为：
[(0.35216413184624645, 'median_income'),
(0.1600064373659322, 'INLAND'),
(0.10973914260172225, 'pop_per_household'),
(0.073196611077795654, 'longitude'),
(0.069993781986234183, 'bedrooms_per_room'),
(0.063532534714522193, 'latitude'),
(0.050927579668676295, 'rooms_per_household'),
(0.043936553336933998, 'housing_median_age'),
(0.015821739675359821, 'total_rooms'),
(0.014974729578331942, 'total_bedrooms'),
(0.014716810447382345, 'households'),
(0.014198191755119289, 'population'),
(0.01174446919617937, '<1H OCEAN'),
(0.0028095259721804206, 'NEAR OCEAN'),
(0.0021732837648423087, 'NEAR BAY'),
(6.4477012541252311e-05, 'ISLAND')]
        这里面0.00几的就不是什么好特征了，应该给删掉，但是因为太麻烦我就不管他了，下面我们要用电脑给我们预测出的参数进行机器学习：

from sklearn.metrics import mean_squared_error
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_pipline.transform(x_test)
final_prediction=final_model.predict(x_test_prepared)

final_mse=mean_squared_error(y_test,final_prediction)
final_rmse=np.sqrt(final_mse)

        然后我们再看看方差：

final_rmse

        我这里显示的是47568.875366719221，又比以前的强了。

freedom100

发现学python  机器学习的坛友很少啊！爬虫的多

fan1993423

不错，有收获。