An ML engineer is analyzing a classification dataset before training a model in Amazon SageMarker AI. The ML engineer suspects that the dataset has a significant imbalance between class labels that…

Question

Options

• AMean square error (MSE)
• BDifference in proportions of labels (DPL)
• CSilhouette score
• DStructural similarity index measure (SSIM)

• A
  3% (1)
• B
  86% (32)
• C
  3% (1)
• D
  8% (3)

Explanation

Difference in Proportions of Labels (DPL) is the standard pre-training bias metric used by Amazon SageMaker Clarify to quantify class imbalance. It measures the difference in the proportion of positive outcomes between a subgroup and the overall dataset, ranging from -1 to +1; values near 0 indicate balance, while values near ±1 indicate severe imbalance. Option A (MSE) is a regression loss metric measuring prediction error, irrelevant before training. Option C (Silhouette score) evaluates the quality of unsupervised clustering results, not class label distribution. Option D (SSIM) is an image quality assessment metric comparing structural similarity between images. Only DPL is a pre-training bias metric designed specifically to detect class label imbalance in classification datasets.

No community discussion yet for this question.