【一起来啃西瓜书】——线性模型

1.基本形式

2.线性回归

1）输入一维，输出一维

2）输入多维，输出一维

3）输入多维，输出多维

4）广义线性模型

3.对数几率回归

4.线性判别分析

5.多分类学习

6.类别不平衡问题

# 导入包
from sklearn.datasets import make_classification
from collections import Counter
from imblearn.over_sampling import RandomOverSampler
# 生成样本集，用于分类算法：3类，5000个样本，特征维度为2
X, y = make_classification(n_samples=5000, n_features=2, n_informative=2,
                           n_redundant=0, n_repeated=0, n_classes=3,
                           n_clusters_per_class=1,
                           weights=[0.01, 0.05, 0.94],
                           class_sep=0.8, random_state=0)

# 打印每个类别样本数
print(Counter(y))

# 过采样
ros = RandomOverSampler(random_state=0)
X_resampled, y_resampled = ros.fit_resample(X, y)

# 打印过采样后每个类别样本数
print(sorted(Counter(y_resampled).items()))

# 导入包
from imblearn.over_sampling import SMOTE

# 过采样
sm = SMOTE(random_state=42)
X_res, y_res = sm.fit_resample(X, y)

# 打印过采样后每个类别样本数
print('Resampled dataset shape %s' % Counter(y_res))

# 导入包
from imblearn.under_sampling import RandomUnderSampler

# 欠采样
rus = RandomUnderSampler(random_state=0)
X_resampled, y_resampled = rus.fit_resample(X, y)

# 打印欠采样后每个类别样本数
print(sorted(Counter(y_resampled).items()))

# 导入包
from imblearn.combine import SMOTEENN
# 过采样与欠采样结合
smote_enn = SMOTEENN(random_state=0)
X_resampled, y_resampled = smote_enn.fit_resample(X, y)
# 打印采样后每个类别样本数
print(sorted(Counter(y_resampled).items()))

# 导入包
from imblearn.combine import SMOTETomek
# 过采样与欠采样结合
smote_tomek = SMOTETomek(random_state=0)
X_resampled, y_resampled = smote_tomek.fit_resample(X, y)
# 打印采样后每个类别样本数
print(sorted(Counter(y_resampled).items()))

# 导入相关包
from sklearn.svm import SVC

# 添加惩罚项
clf = SVC(C=0.8, probability=True, class_weight={0:0.25, 1:0.75})