[github-actions]

✨ FAQ NEW

Course: machine-learning-zoomcamp
Section: module-4 (This is about how to set up cross-validation folds for classification evaluation, which fits the Module 4 section on cross-validation and ROC AUC.)
Related Issue: #466

Question

Homework 4 Q5: For 5-fold cross-validation, should KFold split df_train or df_full_train?

Decision Rationale

No existing FAQ explicitly answers which dataframe to split for 5-fold cross-validation in Homework 4 Q5 (whether to use df_train or df_full_train).

Placement Details

• Section ID: module-4
• Sort Order: End of section
• Filename Slug: homework-4-q5-kfold-split-df-full-train-vs-df-train

⚠️ Warnings

• Order set to -1 (append) because no closely matching existing entry provides an exact placement spot.

---

🤖 Generated by FAQ Bot

Closes #466

[saeidsoheily]

For Homework 4 Q5, call kfold.split(...) on df_full_train, not df_train.

The reason is that cross-validation replaces the usual train/validation split. df_full_train is the full 80% training portion, and 5-fold CV repeatedly splits that same 80% into training and validation folds.

from sklearn.model_selection import KFold

kfold = KFold(n_splits=5, shuffle=True, random_state=1)

for train_idx, val_idx in kfold.split(df_full_train):
    df_train = df_full_train.iloc[train_idx]
    y_train = df_train.converted.values

    df_val = df_full_train.iloc[val_idx]
    y_val = df_val.converted.values

Then separate the target inside each fold, Also fit a new DictVectorizer inside each fold. Note that: Do not fit the vectorizer once outside the CV loop, because that would expose validation-fold categories to training and create leakage.