Examples of Scikit-learn Usages

Examples of Machine Learning Toolkit Usage

Scikit-learn

KFold K-折交叉验证

>>> import numpy as np
>>> from sklearn.model_selection import KFold

>>> X = ["a", "b", "c", "d"]
>>> kf = KFold(n_splits=2)
>>> for train, test in kf.split(X):
...     print("%s %s" % (train, test))
[2 3] [0 1]
[0 1] [2 3]

Reference : http://scikit-learn.org/stable/modules/cross_validation.html#k-fold

Decision Trees Classification 决策树分类

>>> from sklearn import tree
>>> X = [[0, 0], [1, 1]]
>>> Y = [0, 1]
>>> clf = tree.DecisionTreeClassifier()
>>> clf = clf.fit(X, Y)
>>> clf.predict([[2., 2.]])
array([1])

Reference : http://scikit-learn.org/stable/modules/tree.html#classification

KNN k近邻

该算法可以用一句成语来帮助理解：近朱者赤近墨者黑。

from sklearn.neighbors import KNeighborsClassifier

knc = KNeighborsClassifier()
knc.fit(X_train, y_train)
y_pred = knc.predict(X_test)

Logistic Regression 逻辑斯蒂回归

>>> from sklearn.linear_model import LogisticRegression
>>> x_train, x_test, y_train, y_test = train_test_split(iris.data, iris.target, test_size=0.25, random_state=33)
>>> model = LogisticRegression(penalty='l2', random_state=0, solver='newton-cg', multi_class='multinomial')
>>> model = fit(x_train, y_train)
>>> y_pred = model.predict(x_test)

Reference: https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html#sklearn.linear_model.LogisticRegression

Leave One Out 留一法

>>> from sklearn.model_selection import LeaveOneOut

>>> X = [1, 2, 3, 4]
>>> loo = LeaveOneOut()
>>> for train, test in loo.split(X):
...     print("%s %s" % (train, test))
[1 2 3] [0]
[0 2 3] [1]
[0 1 3] [2]
[0 1 2] [3]

Reference : http://scikit-learn.org/stable/modules/cross_validation.html#leave-one-out-loo

train_test_split 随机分割

随机地，将数组或矩阵分割成训练集和测试集

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split

iris = load_iris()
x_train, x_test, y_train, y_test = train_test_split(iris.data, iris.target, test_size=0.25, random_state=33)

参数 test_size

如果是 float，应该在0到1之间，并且代表数据集在列车分割中所包含的比例。

如果是 int，表示训练样本的绝对数量。

如果是 None，则自动将值设置为测试大小的补充。

参数 random_state

如果 int，随机状态是随机数生成器所使用的种子；

如果是 RandomState 实例，随机数是随机数生成器；

如果是 None，随机数生成器是NP-随机使用的随机状态实例。

StandardScaler 特征标准化

标准化数据特征，保证每个维度的特征数据方差为1，均值为0。使得预测结果1不会被某些维度过大的特征而主导

from sklearn.preprocessing import StandardScaler

ss = StandardScaler()
X_train = ss.fit_transform(X_train)
X_test = ss.transform(X_test)

Reference: 《Python机器学习及实践》 https://book.douban.com/subject/26886337

实践

StandardScaler 在鸢尾花（Iris）数据上的表现并不好。未使用 StandardScaler 处理特征时，可以获得：

accuracy 0.947368

avg precision 0.96

avg recall 0.95

f1-score 0.95

代码如下：

# -*- encoding=utf8 -*-

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from sklearn.metrics import classification_report

if __name__ == '__main__':
    iris = load_iris()
    X_train, X_test, y_train, y_test = train_test_split(iris.data, iris.target, test_size=0.25, random_state=33)

    knc = KNeighborsClassifier()
    knc.fit(X_train, y_train)
    y_pred = knc.predict(X_test)

    print("accuracy is %f" % (knc.score(X_test, y_test)))
    print(classification_report(y_test, y_pred, target_names=iris.target_names))

使用了 StandardScaler 以后，这四个指标反而下降了，分别如下所示：

accuracy 0.894737

avg precision 0.92

avg recall 0.89

f1-score 0.90

而使用了 StandardScaler 的代码如下：

# -*- encoding=utf8 -*-

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from sklearn.metrics import classification_report
from sklearn.preprocessing import StandardScaler

if __name__ == '__main__':
    iris = load_iris()
    X_train, X_test, y_train, y_test = train_test_split(iris.data, iris.target, test_size=0.25, random_state=33)

    # 标准化数据特征，保证每个维度的特征数据方差为1，均值为0.
    # 使得预测结果1不会被某些维度过大的特征而主导
    ss = StandardScaler()
    X_train = ss.fit_transform(X_train)
    X_test = ss.transform(X_test)

    knc = KNeighborsClassifier()
    knc.fit(X_train, y_train)
    y_pred = knc.predict(X_test)

    print("accuracy is %f" % (knc.score(X_test, y_test)))
    print(classification_report(y_test, y_pred, target_names=iris.target_names))

这是一个奇怪的问题，需要今后更进一步的探究。

shuffle 随机打乱

该函数可以随机地打乱训练数据和测试数据（让训练数据和测试数据保持对应）

from sklearn.utils import shuffle

x = [1,2,3,4]
y = [1,2,3,4]

x,y = shuffle(x,y)

Out:

x : [1,4,3,2]

y : [1,4,3,2]

Reference : http://scikit-learn.org/stable/modules/generated/sklearn.utils.shuffle.html

Classification Report

Presicion, recall and F1-score.

>>> from sklearn.metrics import classification_report
>>> print(classification_report(y_test, y_pred, target_names=iris.target_names))

              precision    recall  f1-score   support

      setosa       1.00      1.00      1.00         8
  versicolor       0.79      1.00      0.88        11
   virginica       1.00      0.84      0.91        19

    accuracy                           0.92        38
   macro avg       0.93      0.95      0.93        38
weighted avg       0.94      0.92      0.92        38

reference: https://scikit-learn.org/stable/modules/generated/sklearn.metrics.classification_report.html#sklearn.metrics.classification_report

XGBoost

from xgboost import XGBClassifier
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report

if __name__ == '__main__':
    iris = load_iris()
    x_train, x_test, y_train, y_test = train_test_split(iris.data, iris.target)

    xgb = XGBClassifier()
    xgb.fit(x_train, y_train)
    y_pred = xgb.predict(x_test)

    print(classification_report(y_test, y_pred))

实验结果

             precision    recall  f1-score   support

          0       1.00      1.00      1.00        14
          1       0.93      1.00      0.97        14
          2       1.00      0.90      0.95        10

avg / total       0.98      0.97      0.97        38