• 超参数：在算法运行前需要决定的参数
• 模型参数：算法过程中学习的参数
  KNN算法没有模型参数，里面的\(k\)是典型的超参数

KNN的超参数

超参数_\(k\)

\(tips\):

• 若是所搜结果为边界则应该拓宽边界继续搜索

from sklearn import datasets
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
digits = datasets.load_digits()

x = digits.data
y = digits.target

x1, x2, y1, y2 = train_test_split(x, y)

best_score = 0.0
best_k = -1
for k in range(1, 100):
    clf = KNeighborsClassifier(k)
    clf.fit(x1, y1)
    score = clf.score(x2, y2)
    # print(score)
    if score > best_score:
        best_k = k
        best_score = score

print(best_k, best_score)

超参数_权重\(weight\)

将距离的倒数作为权重，当分为\(k\)类，可能会出现平票的现象，使用权重可以有效避免。

如图所示

这样的情况下红色胜出
\(weights\)的参数

• \(uniform\): 不考虑权重
• \(distance\): 距离的倒数的和作为权重

from sklearn import datasets
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
digits = datasets.load_digits()

x = digits.data
y = digits.target

x1, x2, y1, y2 = train_test_split(x, y)

best_score = 0.0
best_k = -1
best_method = ""
for method in ["uniform", "distance"]:
    for k in range(1, 20):
        clf = KNeighborsClassifier(k, weights=method)
        clf.fit(x1, y1)
        score = clf.score(x2, y2)
        # print(score)
        if score > best_score:
            best_k = k
            best_score = score
            best_method = method

print(best_k, best_score,best_method)

超参数_\(p\)

明可夫斯基距离对应的\(p\)

#将前面几个综合起来了
from sklearn import datasets
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
digits = datasets.load_digits()

x = digits.data
y = digits.target

x1, x2, y1, y2 = train_test_split(x, y)

best_score = 0.0
best_k = -1
best_method = ""
best_p = -1
for method in ["uniform", "distance"]:
    for k in range(1, 20):
        for p in range(1, 6):
            clf = KNeighborsClassifier(k, weights=method, p=p)
            clf.fit(x1, y1)
            score = clf.score(x2, y2)
            if score > best_score:
                best_k = k
                best_score = score
                best_method = method
                best_p = p

print(best_k, best_score, best_method, best_p)


#输出结果
1 0.9955555555555555 uniform 4

网格搜索

直接使用\(sklearn\)里面的内置函数来进行测试

from sklearn import datasets
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import GridSearchCV


digits = datasets.load_digits()

x = digits.data
y = digits.target

x1, x2, y1, y2 = train_test_split(x, y)

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)]
    }
]

clf = KNeighborsClassifier()

# 可选 n_jobs 调用多少核去计算， verbose 计算的时候展示信息
grid_search = GridSearchCV(clf, param_grid, n_jobs=-1, verbose=2)

grid_search.fit(x1, y1)

print(grid_search.best_score_)
print(grid_search.best_params_)
print(grid_search.best_estimator_)




# 测试结果
# 过程的一部分
[CV] END ..............n_neighbors=10, p=4, weights=distance; total time=   0.3s
[CV] END ..............n_neighbors=10, p=4, weights=distance; total time=   0.3s
[CV] END ..............n_neighbors=10, p=4, weights=distance; total time=   0.3s
[CV] END ..............n_neighbors=10, p=5, weights=distance; total time=   0.3s
[CV] END ..............n_neighbors=10, p=5, weights=distance; total time=   0.3s
[CV] END ..............n_neighbors=10, p=5, weights=distance; total time=   0.3s
[CV] END ..............n_neighbors=10, p=5, weights=distance; total time=   0.3s
[CV] END ..............n_neighbors=10, p=5, weights=distance; total time=   0.3s
0.989607600165221
{'n_neighbors': 1, 'weights': 'uniform'}
KNeighborsClassifier(n_neighbors=1)