Scikit-learn（sklearn）基本操作（六）

下面完整代码在github仓库：传送门

---

一、Normalizer()

from sklearn import preprocessing
import numpy as np

""""
6.Normalizer()
norm：可以为l1、l2或max，默认为l2
若为l1时，样本各个特征值除以各个特征值的绝对值之和
若为l2时，样本各个特征值除以各个特征值的平方之和
若为max时，样本各个特征值除以样本中特征值最大的值
"""

x = np.array([[1, -1, 2.],
              [2., 0, 0],
              [0, 1, -1]])

scaler = preprocessing.Normalizer(norm="l2")
x_scale = scaler.fit_transform(x)  # 先拟合数据，再标准化
print(x_scale)
print(x_scale.mean(0), x_scale.std(0))


# 数值二值化
scaler = preprocessing.Binarizer(threshold=0)
x_scale = scaler.fit_transform(x)
print(x_scale)
'''
[[1. 0. 1.]
 [1. 0. 0.]
 [0. 1. 0.]]
'''

二、OneHotEncoder()

from sklearn import preprocessing
import numpy as np

# one-hot

enc = preprocessing.OneHotEncoder()
enc1 = preprocessing.OneHotEncoder(sparse=False)

ans = enc.fit_transform([[0], [1], [2], [1]])
ans1 = enc1.fit_transform([[0], [1], [2], [1]])

print(ans)
print(ans1)

'''
  利用numpy实现one-hot编码，比如[5, 2, 8, 6], 

'''

y = np.array([5, 2, 8, 6])

print(np.eye(10)[y])

三、SimpleImputer（）

from sklearn import preprocessing
from sklearn.impute import SimpleImputer
import numpy as np

'''  缺失数据
    preprocessing: ["mean", "median", "most_frequent", "constant"]

'''

# imp = preprocessing.Imputer(missing_values='NaN', strategy='mean')  # 新版该API无法使用
imp = SimpleImputer(missing_values=np.nan, strategy='mean')
y_imp = imp.fit_transform([[np.nan, 2], [6, np.nan], [7, 6]])
print(y_imp)

# 学习填值，使用第一组数的均值来给第二组数填充。
imp.fit([[1, 2], [np.nan, 3], [7, 6]])
y_imp = imp.transform([[np.nan, 2], [6, np.nan], [7, 6]])
print(y_imp)

四、DecisionTreeClassifier()

import numpy as np
from sklearn import tree, datasets, preprocessing
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
import graphviz

# 加载数据
iris = datasets.load_iris()
wine = datasets.load_wine()

# 划分训练集和测试集
x, y = iris.data, iris.target
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.3, random_state=0)
f1 = open("./data/iris.data", 'w', encoding="utf-8")
f2 = open("./data/iris.label", 'w', encoding="utf-8")
f1.write(str(x))
f2.write(str(y))
f1.close()
f2.close()

# 数据预处理
scaler = preprocessing.StandardScaler().fit(x_train)
x_train = scaler.transform(x_train)
x_test = scaler.transform(x_test)

# 创建模型
clf = tree.DecisionTreeClassifier(max_depth=5)

# 模型拟合
clf.fit(x_train, y_train)

# 预测
y_pred = clf.predict(x_test)

# 评估
print(accuracy_score(y_test, y_pred))

# 画决策树图
dot_data = tree.export_graphviz(clf, out_file=None,
                                feature_names=iris.feature_names,
                                class_names=iris.target_names,
                                filled=True, rounded=True,
                                special_characters=True)

graph = graphviz.Source(dot_data)
graph.render("iris")