交叉验证

# knn的算法思想就是看离他最近的几个元组都是属于哪个类，把最多的作为他的类别
from sklearn import datasets
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import cross_val_score
import matplotlib.pyplot as plt

iris = datasets.load_iris()
iris_X = iris.data
iris_y = iris.target

# train_d, test_d, train_tar, test_tar = train_test_split(iris_X, iris_y, test_size=0.2, random_state=5)
# # 随机划分测试集和训练集
#
# knn = KNeighborsClassifier(n_neighbors=5)
# # 寻找离他最近的5个元组并确定每个类别占的比例
# knn.fit(train_d, train_tar)
#
# # 用训练好的函数预测
#
# print('R^2：',knn.score(test_d,test_tar))
# print(knn.predict_proba(test_d))


# 方法二

knn = KNeighborsClassifier(n_neighbors=5)
scores = cross_val_score(knn, iris_X, iris_y, cv=5, scoring='accuracy')
# scoring常用参数：r2,accuracy
# 将数据分成不同的5组，分别计算他的得分（在这里是准确度）
print(scores)

# 选择参数

k_scores = []
k_loss = []
for k in range(1, 31):
    knn = KNeighborsClassifier(n_neighbors=k)
    scores = cross_val_score(knn, iris_X, iris_y, cv=10, scoring='accuracy')
    loss = -cross_val_score(knn, iris_X, iris_y, cv=10, scoring='neg_mean_squared_error')
    # 做回归问题
    k_scores.append(scores.mean())
    k_loss.append(loss.mean())

fig = plt.figure(figsize=(100, 50), dpi=80)

ax1 = fig.add_subplot(111)

ax1.plot(list(range(1, 31)),k_scores, label='accuracy',color='green')
ax2 = ax1.twinx()
ax2.plot(list(range(1, 31)), k_loss, label='loss',color="red")
ax1.legend()
ax2.legend()
ax1.set_ylabel('accuracy')  # 左Y轴的label
ax2.set_ylabel('loss')  # 设置xlabel
plt.show()

# 可以看出13的时候最好