笔记——7.9 GridSearchCV

网格搜索

import numpy as np
from sklearn import datasets

digits=datasets.load_digits()
X=digits.data
y=digits.target

from sklearn.model_selection import train_test_split
X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=0.2,random_state=666)

from sklearn.neighbors import KNeighborsClassifier
knn_clf=KNeighborsClassifier(n_neighbors=4,weights="uniform")
knn_clf.fit(X_train,y_train)
knn_clf.score(X_test,y_test)

0.9916666666666667

Grid Search

import numpy as np
from sklearn import datasets

digits=datasets.load_digits()
X=digits.data
y=digits.target

from sklearn.model_selection import train_test_split
X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=0.2,random_state=666)

param_grid=[
    {
        'weights':['uniform'],
        'n_neighbors':[i for i in range(1,11)]
    },
    {
        'weights':['distance'],
        'n_neighbors':[i for i in range(1,11)],
        'p':[i for i in range(1,6)]
    }
]

实例化一个空的分类器

from sklearn.neighbors import KNeighborsClassifier
knn_clf=KNeighborsClassifier()

n_jobs表示运行内核数，-1表示全部内核都运行，verbose表示输出

from sklearn.model_selection import GridSearchCV
grid_search=GridSearchCV(knn_clf,param_grid,n_jobs=-1,verbose=2)

%%time
grid_search.fit(X_train,y_train)

Fitting 5 folds for each of 60 candidates, totalling 300 fits


[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.
[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:    3.4s
[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:   18.6s


Wall time: 44.7 s


[Parallel(n_jobs=-1)]: Done 300 out of 300 | elapsed:   44.6s finished





GridSearchCV(cv=None, error_score=nan,
             estimator=KNeighborsClassifier(algorithm='auto', leaf_size=30,
                                            metric='minkowski',
                                            metric_params=None, n_jobs=None,
                                            n_neighbors=1, p=2,
                                            weights='uniform'),
             iid='deprecated', n_jobs=-1,
             param_grid=[{'n_neighbors': [1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
                          'weights': ['uniform']},
                         {'n_neighbors': [1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
                          'p': [1, 2, 3, 4, 5], 'weights': ['distance']}],
             pre_dispatch='2*n_jobs', refit=True, return_train_score=False,
             scoring=None, verbose=2)

得到最好的模型best_estimatoe_

grid_search.best_estimator_

KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',
                     metric_params=None, n_jobs=None, n_neighbors=1, p=2,
                     weights='uniform')

最好模型的准确率

grid_search.best_score_

0.9860820751064653

返回最佳参数

grid_search.best_params_

{'n_neighbors': 1, 'weights': 'uniform'}

knn_clf=grid_search.best_estimator_

knn_clf.score(X_test,y_test)

0.9833333333333333

更多距离定义

向量空间余弦相似度	调整余弦相似度	皮尔斯相关系数	Jaccard相似系数
评估两个向量的夹角的相似度，计算两个向量的夹角余弦值，cos=a•b/ % a的模*b的模	调整余弦相似性是通过在计算余弦公式之前减去平均值来实现的	皮尔逊相关系数适用于：(1)、两个变量之间是线性关系，都是连续数据。(2)、两个变量的总体是正态分布，或接近正态的单峰分布。(3)、两个变量的观测值是成对的，每对观测值之间相互独立。	给定两个集合A,B jaccard 系数定义为A与B交集的大小与并集大小的比值，jaccard值越大说明相似度越高，当A和B都为空时，jaccard(A,B)=1

grid_search.fit(X_train,y_train)报错原因

1. weights写成了weight
2. 要重新导入sklearn的包，不然报错，具体原因我也不知道。视频里没报错，但是自己运行却报错

param=[
    {
        'weights':['uniform'],
        'n_neighbors':[i for i in range(1,11)]
    },
    {
        'weights':['distance'],
        'n_neighbors':[i for i in range(1,11)],
        'p':[i for i in range(1,6)]
    }
]