https://www.bilibili.com/video/av29439387
项目背景:信贷数据,是否是一笔正常、合理的贷款
方案:本case采用LR来作检测
1、样本不均衡解决方案:
主要有过采样、下采样
代码示例(下采样):
number_records_fraud = len(data[data.Class==1])
fraud_indeces = np.array(data[data.Class==1].index)
normal_indices = data[data.Class == 0].index
random_normal_indices = np.random.choice(normal_indices,number_records_fraud,replace=True)
random_normal_indices = np.array(random_normal_indices)
under_sample_indices = np.concatenate([fraud_indeces,random_normal_indices])
under_sample_dfata = data.iloc[under_sample_indices,:]
2、交叉验证和模型评估
train_test_split(),注意设置random_state
3、召回率和阈值:可以通过混淆矩阵来观察分布的变化
4、正则化:L1(权重系数的绝对值之和)和L2(权重系数的平方之和),可以遍历可能的C值来调整惩罚力度
5、过采样和SMOTE采样
from imblearn.over_sampling import SMOTE
oversample = SMOTE(random_state=0)
os_features,os_labels = oversample.fit_sample(feature_train,label_train)
生成后的label为1的与label为0的数量一致
与下采样相比,过采样后得出的模型整体精度会高些,召回率会低点。建议采用过采样,数据越多越准确嘛