按标签分隔数据框（将数据框转换为 numpy 数组）

Question

我有一个数据框，我想根据它们的标签将它们分成不同的数组，我不确定如何通过它的索引过滤它。不确定这是否正确完成：

数据集示例（df）

Cancer_Type  | Variable | Data Split | Target
Cancer1         43        Train        Good
Cancer5         34        Train        Bad
Cancer2         34        Test         Good
Cancer3         23        Test         Bad
Cancer4         25        Test         Good

可能会做这样的事情吗？

#initial split into train/test data
train = df['split'] == 'train'
print("train")
print(train)
test = df['split'] == 'test'
print("valid")
print(test)

X_test = test.values[-1, :-1]
y_test = test.values[-1, -1]

# Get the remaining dataset
X = train.values[:-1, :-1]
y = train.values[:-1, -1]

print("X")
#print(type(X))
#print(X)

print("y")
#print(type(y))
#print(y)

# Split the remaining dataset into train and calibration sets.
X_train, X_cal, y_train, y_cal = train_test_split(X, y)
 
print(X_train.shape, y_train.shape)
print(X_cal.shape, y_cal.shape)

希望按行。

Answer 1

据我了解，您希望根据观察值的 Data Split 值将数据拆分为训练集和测试集。之后，您将再次将训练集拆分为训练和校准。标准数据预处理方法包括创建我们的特征 X 和我们的目标 y。

# Get dataframes of train and test features
X_train = df[df['Data Split'] == 'Train'].drop(columns = ['Target']).to_numpy()
X_test = df[df['Data Split'] == 'Test'].drop(columns = ['Target']).to_numpy()

# Get arrays of train and test targets
y_train = df[(df['Data Split'] == 'Train')]["Target"].to_numpy()
y_test = df[(df['Data Split'] == 'Test')]["Target"].to_numpy()

# Split the train dataset further into train and validation/calibration sets.
X_train, X_cal, y_train, y_cal = train_test_split(X_train, y_train)

您现在拥有数组形式的训练、验证/校准和测试集。

如果您希望保留 Target 变量，只需

train = df[df['Data Split'] == 'Train'].to_numpy()
test = df[df['Data Split'] == 'Test'].to_numpy()