数据集划分与特征工程：避免数据泄漏

import pandas as pd
from sklearn.compose import ColumnTransformer
from sklearn.impute import SimpleImputer
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import f1_score
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import OneHotEncoder, StandardScaler

data = pd.DataFrame({
    "input_mb": [10, 15, None, 35, 42, 8, 55, 28, 61, 17, 47, 23],
    "queue_depth": [1, 2, 1, 5, 7, 0, 9, 3, 10, 2, 8, 4],
    "task_type": ["text", "table", "text", "image"] * 3,
    "timeout": [0, 0, 0, 1, 1, 0, 1, 0, 1, 0, 1, 0],
})
X = data.drop(columns="timeout")
y = data["timeout"]
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.33, stratify=y, random_state=42,
)

preprocess = ColumnTransformer([
    ("number", Pipeline([
        ("imputer", SimpleImputer(strategy="median")),
        ("scale", StandardScaler()),
    ]), ["input_mb", "queue_depth"]),
    ("category", OneHotEncoder(handle_unknown="ignore"), ["task_type"]),
])
model = Pipeline([
    ("preprocess", preprocess),
    ("model", LogisticRegression(max_iter=1000)),
])
model.fit(X_train, y_train)
print("test F1:", f1_score(y_test, model.predict(X_test)))
print("new:", model.predict(pd.DataFrame([{
    "input_mb": 30, "queue_depth": 6, "task_type": "archive",
}])).tolist())

from sklearn.model_selection import GroupShuffleSplit

groups = data.index.to_series().map(lambda index: f"device-{index // 3}")
splitter = GroupShuffleSplit(n_splits=1, test_size=0.25, random_state=42)
train_index, test_index = next(splitter.split(X, y, groups=groups))
print("group overlap:", set(groups.iloc[train_index]) & set(groups.iloc[test_index]))

required_columns = {"input_mb", "queue_depth", "task_type"}
incoming = pd.DataFrame([{
    "input_mb": -2, "queue_depth": 6, "task_type": "archive",
}])

missing = required_columns - set(incoming.columns)
if missing:
    raise ValueError(f"missing columns: {sorted(missing)}")
if (incoming["input_mb"].dropna() < 0).any():
    raise ValueError("input_mb must be non-negative")