AWS Certified AI Practitioner (AIF-C01) free question 1 / 10 · Domain 1: Fundamentals of AI and ML
A retailer has purchase histories for its members, but no one has an answer for how the members should be segmented. The retailer wants to group members who buy in similar ways and vary promotions by group. According to the Amazon SageMaker AI built-in algorithms page, which technique should it use?
- Clustering of similar members with K-Means
- Regression to predict purchase amounts
- Dimensionality reduction of columns with PCA
- Multiclass classification into set segments
Show the answer and explanation
Answer: A
Why this is correct
The built-in algorithms page treats clustering (grouping) as the problem of grouping similar objects or data. Its example is finding customers with high, medium, and low spending from transaction history, and it shows K-Means as the algorithm. It sits in the unsupervised learning section and can find groups in data that has no answer for the segments. That matches a case where nobody holds the correct segmentation.
Why the other options are wrong
"Multiclass classification into set segments" … Multiclass classification is supervised learning that assigns items to predefined categories (on the page, text classification assigns predefined categories to documents in a corpus). It cannot learn under the question's condition that no segment answers exist.
"Regression to predict purchase amounts" … Regression is supervised learning that predicts numeric or continuous values, such as the price of a house. It cannot divide members into groups.
"Dimensionality reduction of columns with PCA" … Dimensionality reduction with PCA is also unsupervised, but the page's example is feature engineering that drops columns that are weakly related to the target variable. It is not a technique for gathering members into groups.
Key point
The unsupervised learning section of the same page lists clustering, dimensionality reduction, pattern recognition, and anomaly detection as unsupervised tasks. Grouping unlabeled data is clustering, reducing columns is dimensionality reduction, and finding outliers is anomaly detection.