When converting data categories before training an AI model, which of the following scenarios represents the GREATEST risk?

Question

Options

• AOne-hot encoding the data attribute car colors for the options red, blue, green, black, white
• BCreating dummy variables for the data attribute dog breed for the options labrador, terrier, beagle
• COne-hot encoding the data attribute customer rewards category for the options economy,
• DCreating dummy variables for the data attribute product flavor for the options vanilla, chocolate,

• A
  9% (2)
• B
  13% (3)
• C
  74% (17)
• D
  4% (1)

Explanation

The greatest risk in Option C is applying one-hot encoding to ordinal data. Customer rewards categories (e.g., economy, standard, premium, gold) have a natural, meaningful ranking-each tier is 'greater than' the one below it. One-hot encoding treats all categories as equally unrelated nominal values, completely destroying the ordinal relationship. This causes the model to lose critical information about tier hierarchy, leading to poor or misleading predictions. The correct approach for ordered categories is ordinal encoding, which preserves the rank. By contrast, car colors (A) and dog breeds (B) are genuinely nominal (no inherent order), making one-hot encoding appropriate. Option C applies the wrong encoding type to the wrong data type, which is the most impactful error.

No community discussion yet for this question.