Free AWS AIP-C01 Actual Exam Questions - Question 10 Discussion

I’m pretty sure it’s C here. DeepAR is made to handle multiple related time series, so it can learn from existing SKUs and forecast for a new variant even with no direct sales data. Clustering in D might group SKUs but can’t actually predict sales numbers on its own. A and B don’t seem to fit since RCF is for anomaly detection and Linear Learner needs historical labels, which we don’t have for the new product. So, C makes the most sense for the cold-start problem.

C/D? Clustering (D) helps find similar products but doesn't directly forecast demand, while DeepAR (C) is specifically made for time series and can leverage patterns across SKUs to predict new variants.

C/D? DeepAR in C is built for time series and can share patterns across SKUs, but D’s clustering could at least group similar SKUs to help infer patterns when no direct sales data exists. A and B seem less suitable.

C DeepAR is designed for time series forecasting across related SKUs by learning shared patterns, so it can handle new variants by leveraging data from existing ones, unlike clustering which doesn’t directly predict sales.

D imo, clustering at least groups similar SKUs for pattern inference without needing sales history.

I’m thinking B could be ruled out since Linear Learner is more for straightforward regression and might not capture the time series patterns across SKUs like DeepAR does. RCF in A is mainly for anomaly detection, so that doesn’t quite fit forecasting. D’s clustering might help group SKUs but doesn’t directly forecast sales. So C still seems best as it handles time series and learns shared patterns across products.

It’s C, since DeepAR captures time series patterns across SKUs, unlike clustering in D.

Maybe D could work too—grouping similar SKUs might give a rough idea for the new variant’s demand by analogy. But it won’t capture time series trends as well as C, so probably less accurate for forecasting.

It’s C because DeepAR models time series well and shares info across SKUs.

Agree, C fits best since it handles multiple SKUs and cold-start forecasting.

Option B seems off since linear regression won’t capture time series patterns well for forecasting. DeepAR (C) is better because it models time dependencies and shares info across SKUs.

This one feels like C makes the most sense. DeepAR is designed to handle multiple related time series, so it can learn from existing SKUs and forecast the new variant even without its own history. A and D don’t really fit because RCF is for anomaly detection, and K-means is more about grouping, not forecasting. B might work but Linear Learner won’t capture complex patterns across SKUs as well as DeepAR. So, C it is.

D looks best here—using AWS KMS for encryption and Glue for redacting credit card numbers before training fits the requirements well.