Data balancing addresses data imbalance by increasing minority class samples; which approach is commonly used?

• A. Increasing majority class samples
• B. Oversampling the minority class
• C. Reducing overall dataset
• D. Ignoring the issue

Answer: (B)

Balancing data helps a model learn from all classes instead of leaning toward the majority. The way to address imbalance by increasing minority class samples is to oversample that class. By adding more minority examples—whether through duplicating existing instances or generating new synthetic ones—the model sees more varied instances of the minority class during training, which improves its ability to recognize and correctly classify those cases. This approach preserves the information from the majority class while giving the minority class more representation, making the decision boundary between classes clearer.

While simple replication can risk overfitting, more advanced oversampling methods (like creating synthetic minority samples) reduce that risk and are widely used in practice. Increasing the majority class would worsen imbalance, reducing the dataset trims information, and ignoring the issue leaves the model biased.