Transaction features are paired with human-verified fraud or legitimate labels. Predict that outcome for new transactions.
A separate purchase dataset has no target customer-segment labels. Explore groups of similar behavior without treating a cluster number as verified fraud.
Choose TWO appropriate mappings between the goal and learning method.
Match labeled outcome prediction to supervised learning and unlabeled group discovery to unsupervised learning. Do not equate a cluster with verified fraud or training performance with guaranteed accuracy on new data.
Detailed explanation
The model learns a mapping from input examples to target labels, matching the prediction goal and labeled training data.
A similarity group is not a verified fraud label. Its features and relationship to fraud need separate investigation.
Training fit and generalization to unseen data differ. Evaluate with appropriate data that was not used for learning.
This explores structure rather than predicting a supplied answer class. The business meaning of each group still needs interpretation.
The distinction concerns target labels used in learning, not whether a person prepared the data. Learning from labeled input-output pairs is supervised.
Check official documentation
Compare the following points with the official documentation.
Compare AWS learning-method definitions with each example input, target labels and requested output. Treat unseen-data evaluation of a classifier and business interpretation of clusters as separate checks.
Official references
What to confirm
Explain labeled transaction classification as supervised and unlabeled customer grouping as unsupervised. Do not infer verified fraud from cluster numbers or guarantee unseen-data accuracy from training results alone.
Key points
- Target labels
- Classification
- Clustering
- Unseen data
Notes
- Environment: Official AIF-C01 Task 1.1 and AI/ML documentation review; no model training or production changes are performed.
- Compare the definitions with current AIF-C01 Task 1.1 and the linked primary sources.
- A conceptual example does not establish measured model accuracy or authorize external operations.
Foundation review
Match the goal to training data
Check both the requested output and the target labels supplied during learning. This scenario contrasts verified fraud classification with customer grouping without segment labels.
Interpret clusters separately
Similar purchase behavior is not automatically a business category or verified fraud. Inspect group characteristics and their meaning before drawing conclusions.
Separate training fit from generalization
Strong training results do not guarantee accuracy on new transactions. Check label quality, representativeness and evaluation data kept separate from learning.