You need to implement a model development strategy to determine a user's tendency to respond to an ad. Which technique should you use?

Question

Options

• AUse a Relative Expression Split module to partition the data based on centroid distance.
• BUse a Relative Expression Split module to partition the data based on distance travelled to the
• CUse a Split Rows module to partition the data based on distance travelled to the event.
• DUse a Split Rows module to partition the data based on centroid distance.

• A
  79% (19)
• B
  4% (1)
• C
  13% (3)
• D
  4% (1)

Explanation

For propensity-to-respond modeling, the approach is to first cluster users and then use centroid distance as a feature - users closer to a cluster of known responders have a higher propensity. In Azure Machine Learning Studio, the Relative Expression Split module partitions data based on a numeric relational expression, making it suitable for splitting data based on a calculated centroid distance value. This creates train/test sets that reflect the distance-based feature distribution. Options B and C reference 'distance travelled to the event', which is a physical/geographic metric unrelated to behavioral propensity scoring. The Split Rows module (C, D) performs row-level splits (e.g., random or stratified), not expression-based splits tied to a computed distance metric like centroid distance.

No community discussion yet for this question.