使用 ADASYN 算法对多类数据进行过采样失败

Question

我下面有一个非常基本的脚本来演示该问题：from imblearn.over_sampling import ADASYNimport pandas as pd, numpy as npfrom sklearn.ensemble import RandomForestClassifierfrom sklearn.model_selection import train_test_splitdata = pd.read_csv('glass.csv')classes = data.values[:, -1]data = data.iloc[:, :-1]adasyn = ADASYN(sampling_strategy='not majority', random_state=8, n_neighbors=3)new_data, new_classes = adasyn.fit_resample(data, classes)X_train, X_test, y_train, y_test = train_test_split(new_data, new_classes, test_size = 0.20)rfc = RandomForestClassifier()rfc.fit(X_train, y_train)print("Score: {}".format(rfc.score(X_test, y_test)))目的是平衡以下类别的不平衡：(214, 10) Class=1, Count=70, Percentage=32.710% Class=2, Count=76, Percentage=35.514% Class=3, Count=17, Percentage=7.944% Class=5, Count=13, Percentage=6.075% Class=6, Count=9, Percentage=4.206% Class=7, Count=29, Percentage=13.551%拥有相等（或接近相等）的样本。然而，运行上面的代码会产生：ValueError: No samples will be generated with the provided ratio settings.更改为成功地对类 进行过采样ADASYN，并将其带入样本，但仍然使其余类不平衡。因此，我正在寻找一种使用 ADASYN对所有少数类别进行完全过采样的方法。sampling_strategyminorityminority674ADASYN 文档指出： 'not majority': resample all classes but the majority class;但这显然没有发生。

Answer 1

为了解决这个问题，我所做的是对除两个主要多数类之外的所有类进行重新采样，并继续通过以下方式进行：

adasyn = ADASYN(sampling_strategy='minority', random_state=8, n_neighbors=3)

new_data = data

new_classes = classes

for i in range(len(classes)-2):

    new_data, new_classes = adasyn.fit_resample(data, classes)