[marcos-Nieves-24]

New FAQ entry for machine-learning-zoomcamp / module-4-homework.

Students hitting Q5 with the bare LogisticRegression(max_iter=1000) constructor get std 0.013 (lbfgs default + ConvergenceWarning) instead of the correct 0.007 (solver='liblinear'). Fix-first answer with runnable per-fold code, verified against the pinned 2026 homework: https://github.com/DataTalksClub/machine-learning-zoomcamp/blob/main/cohorts/2026/homework/04-evaluation/homework.md

[saeidsoheily]

You should explicitly set solver='liblinear' when creating your Logistic Regression model (The default solver is lbfgs).

# Incorrect: uses the default lbfgs solver
model = LogisticRegression(max_iter=1000)

# Correct: required by the 2026 homework
model = LogisticRegression(
    solver='liblinear',
    C=1.0,
    max_iter=1000
)

The [official ML Zoomcamp 2026 Module 4 homework] explicitly requires solver='liblinear', C=1.0, and max_iter=1000 for Question 5.

The difference matters because lbfgs and liblinear use different optimization algorithms. On this dataset, they can produce different model coefficients, predictions, and ROC AUC scores.