AWS MLS-C01 Free Practice Questions — Page 1

Correct Answer: C. The model is 86% accurate and the cost incurred by the company as a result of false positives isfiless than the false negatives.

Why Option C is right: In this specific business context, the company states that the cost of churn is far greater than the cost of the incentive. A False Negative (Actual Churn "Yes", Predicted "No") means a customer leaves, which is the highest cost. A False Positive (Actual Churn "No", Predicted "Yes") means giving an incentive to a loyal customer. While this is a small loss, it is much better than losing the customer. Since the model captures most churners and maintains high overall accuracy, it is a viable business solution.

Correct Answer: B. Build a collaborative filtering recommendation engine with Apache Spark ML on Amazon EMR.

Collaborative filtering is the correct approach for this use case because it predicts user preferences based on similarities between users, not on product content characteristics. The question explicitly states the objective is to "predict which products users would like based on the users' similarity to other users," which is the defining principle of collaborative filtering. Apache Spark ML on Amazon EMR is an appropriate implementation platform for this algorithm, as it provides distributed computing capabilities for large-scale data processing. Collaborative filtering works by identifying user-user or item-item similarities and leveraging the preferences of similar users to make recommendations.Why correct: Collaborative filtering explicitly uses user similarity to make predictions Aligns with stated objective: "based on the users' similarity to other users" Spark ML on EMR supports distributed collaborative filtering implementation Why others are incorrect: A (Content-based filtering): Uses product features/content, not user similarity C (Model-based filtering): A variant of collaborative filtering, but less direct for this use case; the question emphasizes similarity between users D (Combinative filtering): Not a standard ML terminology; likely a distractor

Correct Answer: D. Ingest .CSV data from Amazon Kinesis Data Streams and use Amazon Kinesis Data Firehose to convert data into Parquet.

Amazon Kinesis Data Firehose with Parquet conversion is the LEAST effort solution because it's a fully managed, serverless service that handles data ingestion, transformation, and S3 delivery without requiring infrastructure management. Firehose can automatically convert incoming CSV data to Parquet format and deliver it directly to S3, eliminating the need to write custom transformation code or manage compute clusters. This approach reduces operational overhead compared to alternatives that require setting up Kafka clusters, EMR clusters, or custom Glue jobs. Firehose integrates natively with S3 and supports data format conversion as a built-in feature, making it the most straightforward implementation path. Why correct: Fully managed, serverless service—no infrastructure to manage Built-in CSV-to-Parquet conversion capability Automatic delivery to S3 Minimal code/configuration required Why others are incorrect: A (Kafka Streams + Kafka Connect): Requires setting up and managing Kafka infrastructure on EC2; significant operational overhead B (Kinesis Data Streams + Glue): Requires Glue job configuration and custom transformation logic; more effort than Firehose C (EMR + Spark Structured Streaming): Requires launching and managing an EMR cluster; highest operational complexity

Correct Answer: C. Use the Amazon SageMaker Linear Learner algorithm on the single time series consisting of the full year of data with a predictor_type of regressor.

Since the objective is to predict a continuous value (parts per million of contaminants), this is a regression task. The Linear Learner algorithm with predictor_type='regressor' is the most appropriate SageMaker built-in algorithm for forecasting a single time series when limited data is available. Option A (kNN) is typically less effective for simple time-series forecasting compared to Linear Learner, Option B (RCF) is for anomaly detection, and Option D (classifier) is for predicting discrete categories rather than continuous numbers.

Correct Answer: D. Use AWS KMS to encrypt the data on Amazon S3 and Amazon SageMaker, and redact the credit card numbers from the customer data with AWS Glue.

Using AWS KMS to encrypt data on both S3 and SageMaker, combined with AWS Glue to redact credit card numbers, is the most secure and appropriate solution for handling sensitive payment data. AWS KMS provides enterprise-grade encryption key management with audit trails, encryption at rest, and encryption in transit—meeting compliance requirements for credit card information (PCI-DSS). AWS Glue's data catalog and ETL capabilities can automatically identify and redact sensitive data patterns (such as credit card numbers) before the data reaches SageMaker. This approach provides defense-in-depth: encryption for confidentiality and data redaction for further protection. KMS keys are managed separately from the data, preventing unauthorized access even if storage is compromised. Why correct: AWS KMS is the standard for sensitive data encryption (PCI-DSS compliant) Redaction removes sensitive data entirely—defense-in-depth Encrypts both at-rest (S3, SageMaker) and in-transit KMS audit logs track all key access Glue can identify sensitive patterns and redact programmatically