A SAS analyst uses PROC HPCLUS to cluster 1 million customers. After obtaining cluster assignments, they want to score new customers arriving each week using the cluster centroids. Which approach is most appropriate?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — you score new customers by applying the saved cluster centroids, either with PROC HPCLUS's SCORE statement or by exporting centroids and assigning each new customer to the nearest centroid in a DATA step.
Full Explanation
The correct approach is to use the SCORE statement in PROC HPCLUS or export centroids and compute distance to each centroid in a DATA step. PROC HPCLUS can score new observations by applying saved centroids, assigning each customer to the nearest centroid. Alternatively, exported centroids let a DATA step compute Euclidean distances and assign clusters. Re-running PROC HPCLUS on combined old and new data weekly is wasteful and changes cluster definitions. Applying PROC HPLOGISTIC to predict cluster membership converts an unsupervised problem into a supervised one unnecessarily. Using PROC DISTANCE for all pairwise distances between new and existing customers is impractical for 1 million existing customers.