机器学习--逻辑回归

逻辑回归：用线性回归式子作为逻辑回归的输入，用来解决二分类问题

想把线性回归 用来做二分类问题，要用sigmoid函数

小于0.5的概率归为0，大于0.5的概率归位1

二、用逻辑回归做癌症二分类问题

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import  SGDRegressor, LogisticRegression
from sklearn.metrics import classification_report    #召回率API
def logistic():
    """
    逻辑回归对癌症病情的二分类问题
    :return: None
    """
    #构造类标签名字
    column = ['Sample code number', 'Clump Thickness',
              'Uniformity of Cell Size',
              'Uniformity of Cell Shape',
              'Marginal Adhesion',
             'Single Epithelial Cell Size ',
              'Bare Nuclei',
              'Bland Chromatin',
              'Normal Nucleoli',
              'Mitoses',
              'Class:']
    data = pd.read_csv("./breast-cancer-wisconsin.data ", names=column)
    print(data)

    #处理缺失值
    data = data.replace(to_replace='?', value=np.nan)
    data = data.dropna()    #把nan类型的值删除

    #进行数据的分割
    x_train, x_test, y_train, y_test = train_test_split(data[column[1:10]], data[column[10]], test_size=0.25)  #切片时 不包括10
    print(x_train )

    #进行标准化处理
    std = StandardScaler()
    x_train = std.fit_transform(x_train)
    x_test = std.fit_transform(x_test)

    #进行逻辑回归预测
    LR = LogisticRegression(C=1.0, )
    LR.fit(x_train, y_train )
    print(LR.coef_)

    y_predict = LR.predict(x_test)

    print("准确率", LR.score(x_test, y_test) )

    print("召回率", classification_report(y_test, y_predict, labels=[2, 4], target_names=['良性', '恶性'] ) )


    return None
if "__name__" =="__name__":
    logistic()

运行结果

   Sample code number  Clump Thickness  ...  Mitoses  Class:
0               1000025                5  ...        1       2
1               1002945                5  ...        1       2
2               1015425                3  ...        1       2
3               1016277                6  ...        1       2
4               1017023                4  ...        1       2
..                  ...              ...  ...      ...     ...
694              776715                3  ...        1       2
695              841769                2  ...        1       2
696              888820                5  ...        2       4
697              897471                4  ...        1       4
698              897471                4  ...        1       4

[699 rows x 11 columns]
     Clump Thickness  Uniformity of Cell Size  ...  Normal Nucleoli  Mitoses
146                3                        4  ...                1        1
426                5                        3  ...                1        1
620                3                        1  ...                1        1
214               10                       10  ...                6        1
336                6                        5  ...                4        1
..               ...                      ...  ...              ...      ...
91                 3                        1  ...                1        1
121                4                        2  ...                1        1
421               10                       10  ...                2        1
311                1                        1  ...                1        1
344                7                        6  ...                5        3

[512 rows x 9 columns]
[[ 1.49940035 -0.03587566  1.04958732  0.50454901  0.58707274  1.36663014
   0.78557885  0.41617397  0.65672664]]
准确率 0.9532163742690059
召回率               precision    recall  f1-score   support

          良性       0.97      0.96      0.97       116
          恶性       0.91      0.95      0.93        55

    accuracy                           0.95       171
   macro avg       0.94      0.95      0.95       171
weighted avg       0.95      0.95      0.95       171

主要关注召回率，看恶性是0.95 ， 意味着假如100个人，有5个人患癌症没有被预测出来