Class Imbalance — Google Cloud ML Engineer Practice Questions

Class imbalance occurs when one or more target classes are significantly underrepresented in training data, causing models to be biased toward the majority class and perform poorly on rare but often important categories. The Google Cloud ML Engineer exam tests strategies for addressing imbalance, including oversampling minority classes, undersampling the majority, generating synthetic samples with techniques like SMOTE, and adjusting class weights during training. Evaluation considerations are also tested: accuracy alone is misleading for imbalanced datasets, so candidates must understand metrics like precision, recall, F1-score, and area under the precision-recall curve. Selecting the right resampling approach depends on dataset size, the cost of false negatives versus false positives, and the specific model being trained.

Free questions on class imbalance

More class imbalance questions in the full bank

• Which metric is most appropriate for evaluating a binary classification model with imbalanced classes? Unlock answer & explanation →
• Your training dataset has a significant class imbalance (95% negative, 5% positive). Which metric should you primarily use to evaluate your binary classification model? Unlock answer & explanation →
• You have an imbalanced dataset with class ratio 1:100. Which combination would be most effective? Unlock answer & explanation →