MLS-C01 Practice Exam 66 — 20 Questions

Question 1

A company is using Amazon Kinesis Data Firehose to deliver streaming data to an S3 bucket. The data is JSON and must be transformed into Parquet format before delivery. Which approach should the data engineer use?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Configure Kinesis Data Firehose to convert the record format to Parquet using a schema from AWS Glue Data Catalog. Kinesis Data Firehose natively supports record format conversion from JSON to Parquet (and ORC) using a schema stored in the AWS Glue Data Catalog. You enable this in the Firehose stream configuration, and Firehose handles the conversion and delivery to S3 without custom code. This is the lowest-effort, fully managed approach.

Answer

Send the data to Amazon Kinesis Data Analytics to convert to Parquet

Answer

Use an AWS Lambda function to transform JSON to Parquet and write to S3

Answer

Use an AWS Glue ETL job to read from Firehose and write Parquet to S3

Question 2

Which TWO of the following are appropriate techniques for handling missing data during exploratory data analysis? (Select TWO.)

Accepted Answer

Impute missing values using mean or median for numerical features. Option C is correct because imputing missing numerical values with the mean or median is a standard, well-established EDA technique that preserves sample size and keeps distributions usable for analysis, with the median being preferable when the data is skewed or contains outliers. Option D is correct because visualizing missingness patterns with heatmaps (e.g., missingno's matrix or heatmap) or bar charts of missing-value counts is a fundamental EDA step that reveals whether missingness is random, systematic, or concentrated in particular features, guiding the choice of imputation or exclusion strategy. Option A is not appropriate because ignoring missing values silently biases results and many algorithms cannot handle NaN inputs at all. Option B is not appropriate because substituting -1 conflates a sentinel value with real data, distorts distributions and statistics, and can mislead distance-based or linear models. Option E is not appropriate because deleting every row with any missing value causes substantial data loss, reduces statistical power, and can introduce bias unless the missingness is provably MCAR and the proportion is tiny.

Answer

Ignore missing values and proceed with modeling

Answer

Replace missing values with -1 to indicate missing

Answer

Delete all rows with any missing values

Question 3

A data scientist is performing EDA on a dataset with a binary target variable. Which THREE techniques can help assess the relationship between a continuous feature and the target?

Accepted Answer

KDE plot grouped by target. Option B (KDE plot grouped by target) is correct because kernel density estimation overlays the distribution of the continuous feature for each class of the binary target, revealing shifts, overlaps, or separation that indicate the strength of the relationship. Option C (Histogram colored by target) is correct because binning the continuous feature and stacking/coloring counts by target class shows how the class-conditional distributions differ across the feature's range. Option E (Box plot grouped by target) is correct because it summarizes each class's median, IQR, and outliers for the continuous feature, making between-group differences easy to compare. Option A is not appropriate because a scatter plot of the continuous feature against another continuous feature does not involve the binary target, so it cannot assess the feature-target relationship. Option D is not appropriate because a bar chart of feature values is suited to categorical or discrete data and does not meaningfully display the distribution of a continuous feature relative to a binary target.

Answer

Scatter plot against another continuous feature

Answer

Bar chart of feature values

Question 4

A data engineering team is designing a data lake on AWS for machine learning workloads. The data includes structured, semi-structured, and unstructured data. The team needs to ensure that the data is cataloged, easily discoverable, and can be queried by Amazon Athena and Amazon EMR. The team also wants to enforce fine-grained access control at the column and row level for sensitive data. Which combination of AWS services should the team use? (Select TWO.)

Accepted Answer

AWS Lake Formation. AWS Lake Formation (A) is correct because it is the service that provides centralized fine-grained access control for data lakes, allowing column-level and row-level security policies to be enforced across analytics services such as Amazon Athena and Amazon EMR. AWS Glue Data Catalog (C) is correct because it serves as the persistent metadata store that catalogs structured, semi-structured, and unstructured data, making the data discoverable and queryable by Athena and EMR. Together, Lake Formation governs permissions while the Glue Data Catalog provides the metadata backbone for the data lake. IAM (B) handles authentication and broad AWS resource permissions but does not by itself deliver column- and row-level data lake access control. Amazon RDS for PostgreSQL (D) is a relational database engine, not a data lake catalog or governance service. Amazon DynamoDB (E) is a NoSQL key-value store and does not provide data lake cataloging or fine-grained analytics access control.

Answer

AWS Identity and Access Management (IAM)

Answer

Amazon RDS for PostgreSQL

Answer

Amazon DynamoDB

Question 5

A data analyst is performing exploratory data analysis on a dataset and notices that there are outliers in several numerical columns. Which TWO methods can the analyst use to identify outliers?

Accepted Answer

Calculate z-scores and flag any data points with |z| > 3.. Option B is correct because calculating z-scores standardizes each value relative to the column's mean and standard deviation, and flagging points with |z| > 3 is a standard statistical rule for identifying extreme outliers in numerical data. Option C is correct because a box plot displays the interquartile range and whiskers (typically 1.5×IQR beyond Q1 and Q3), so points plotted beyond the whiskers are explicitly identified as outliers. Option A is not a reliable outlier-identification method here, since a scatter plot matrix shows pairwise relationships and may not clearly reveal univariate outliers in several numerical columns. Option D does not identify outliers; comparing mean and median only indicates skewness or asymmetry in a distribution. Option E is also insufficient, as histogram gaps may reflect binning choices or multimodal distributions rather than true outliers.

Answer

Create a scatter plot matrix to visually inspect.

Answer

Compare the mean and median of each column.

Answer

Plot a histogram and look for gaps.

Question 6

A data scientist is exploring a dataset with 200 features. They compute the pairwise correlation matrix and notice that many features have correlations above 0.95. They want to reduce redundancy before modeling. Which of the following techniques is most appropriate for identifying and removing highly correlated features?

Accepted Answer

Perform hierarchical clustering on the correlation matrix and select one feature per cluster.. Hierarchical clustering on correlations groups correlated features; then one can select a representative from each cluster. Option A is wrong because mutual information with the target does not capture pairwise redundancy among features. Option B is wrong because PCA creates new features but does not remove original ones. Option C is wrong because Lasso regression performs feature selection but may not handle multicollinearity well and does not directly identify redundant groups.

Answer

Compute mutual information between each feature and the target.

Answer

Apply PCA and keep the first 50 components.

Answer

Use Lasso regression to select features.

Question 7

A machine learning engineer needs to process a large dataset that does not fit on a single Amazon SageMaker notebook instance's EBS volume. The data is stored in S3. What is the MOST efficient way to access the data from the notebook?

Accepted Answer

Read the data directly from S3 using the boto3 library.. Reading data directly from S3 using the boto3 library is the most efficient approach for a dataset that exceeds the notebook instance's EBS volume capacity. Boto3 allows you to stream data in chunks or use S3 Select for server-side filtering, avoiding the need to download the entire dataset to local storage. This method leverages S3's high-throughput API and eliminates the bottleneck of writing to a local EBS volume, which is limited in size and I/O performance.

Answer

Increase the EBS volume size to 5 TB.

Answer

Mount the S3 bucket as a file system using s3fs.

Answer

Use SageMaker File input mode in the notebook.

Question 8

Which TWO data formats are columnar and optimized for analytics queries in Amazon S3?

Accepted Answer

ORC. ORC (B) is a columnar storage format that stores data by column rather than row, enabling predicate pushdown, column pruning, and compression that make analytics queries in Amazon S3 (e.g., via Athena, EMR, or Redshift Spectrum) far more efficient. Parquet (E) is likewise a columnar format with schema evolution, efficient encoding, and compression, and is natively supported by AWS analytics services for high-performance S3 queries. CSV (A) is a row-based plain-text format with no columnar layout or built-in compression, so it is not optimized for analytics. JSON (C) is a row-oriented, self-describing text format that is flexible for ingestion but inefficient for large-scale analytical scans. Avro (D) is a row-based format designed for serialization and streaming/write-heavy workloads, not columnar analytics.

Answer

CSV

Answer

JSON

Answer

Avro

Question 9

A company uses Amazon SageMaker to deploy a real-time inference endpoint for a regression model. The endpoint is experiencing high latency during spikes in traffic. The data scientist needs to reduce latency while maintaining cost efficiency. Which action should the data scientist take?

Accepted Answer

Enable automatic scaling for the endpoint. Enabling automatic scaling for the SageMaker endpoint allows the number of instances to dynamically adjust based on traffic patterns, reducing latency during spikes by adding capacity when needed and removing it during low traffic to maintain cost efficiency. Automatic scaling uses CloudWatch metrics (e.g., InvocationsPerInstance or CPUUtilization) to trigger scale-out and scale-in policies, ensuring the endpoint can handle bursts without over-provisioning.

Answer

Use batch transform instead of real-time inference

Answer

Use a larger instance type for the endpoint

Answer

Deploy the model on a multi-model endpoint

Question 10

Which TWO options are best practices for managing access to data stored in Amazon S3 for a data lake?

Accepted Answer

Use IAM policies to control user and role permissions. Option C is correct because IAM policies are the primary, centrally managed mechanism for granting users and roles fine-grained permissions to S3 data lake resources, allowing least-privilege access tied to identities. Option D is correct because S3 bucket policies are resource-based policies that can explicitly grant or deny cross-account access, which is essential when external accounts or partners need controlled access to the data lake. Option A is not a best practice because S3 ACLs are legacy, coarse-grained, and AWS recommends disabling them in favor of IAM and bucket policies. Option B is not a best practice for access management because SSE-S3 default encryption protects data at rest but does not control who can access it. Option E is not a best practice for general access management because pre-signed URLs grant temporary access to specific objects and are unsuitable as a primary, scalable authorization strategy.

Answer

Use S3 access control lists (ACLs) for granular permissions

Answer

Enable default encryption with SSE-S3

Answer

Generate pre-signed URLs for all data access

Question 11

A team is building a data pipeline to process terabytes of log data daily using Amazon EMR. The data arrives in 5-minute windows and must be available for querying within 30 minutes. The data is originally in gzip-compressed CSV files. Which approach will minimize processing time and cost?

Accepted Answer

Use Amazon EMR with Spark to convert data to Parquet and store in S3, using spot instances for task nodes.. Converting gzip-compressed CSV to Parquet reduces storage size and improves query performance due to columnar storage and predicate pushdown. Using spot instances for task nodes significantly lowers compute cost, while the 30-minute SLA is achievable with Spark on EMR processing 5-minute windows of data.

Answer

Use Amazon EMR with Spark to convert data to Parquet and use on-demand instances.

Answer

Use AWS Glue to convert data to gzip-compressed CSV and query with Athena.

Answer

Use Amazon EMR with Hive to transform data to compressed CSV and store in S3.

Question 12

Refer to the exhibit. A data scientist is setting up an IAM policy for EDA on a data lake. The scientist needs to run exploratory SQL queries using Amazon Athena and save results to a new S3 bucket. What is a critical missing permission in this policy?

Accepted Answer

s3:PutObject on the output S3 bucket. Athena writes query results to an S3 output bucket, which requires the s3:PutObject permission. Option A is incorrect because s3:ListBucket on the output bucket is not needed to write results; s3:PutObject is the necessary write permission. Option C is incorrect because glue:GetDatabase is not required for the query execution; the policy likely already includes glue:GetTable. Option D is incorrect because athena:StopQueryExecution is not directly related to saving results; the missing permission for saving results is s3:PutObject.

Answer

s3:ListBucket on the output bucket

Answer

glue:GetDatabase

Answer

athena:StopQueryExecution

Question 13

Which TWO of the following are benefits of using SageMaker Managed Spot Training? (Select TWO.)

Accepted Answer

Potential for significant cost savings. Option B is correct because SageMaker Managed Spot Training uses spare EC2 capacity that AWS prices at a discount of up to 90% versus On-Demand, so workloads can achieve significant cost savings. Option E is correct because the same underlying mechanism means the effective training cost is lower than running equivalent On-Demand instances, which is the primary reason to adopt Managed Spot Training. Note that these two benefits are closely related: the discount on spare capacity is what produces the lower cost relative to On-Demand. Option A is not correct because Spot capacity can be reclaimed with a two-minute interruption notice, so checkpointing to Amazon S3 is strongly recommended (and often required) to resume training. Option C is not correct because Spot instances are not inherently faster; training time depends on instance type and workload, and interruptions can actually lengthen wall-clock time. Option D is not correct because Spot capacity is not guaranteed—AWS can reclaim it when capacity is needed, which is precisely why Managed Spot Training must handle interruptions.

Answer

No need to checkpoint the model

Answer

Faster training times

Answer

Guaranteed instance availability

Question 14

A data scientist is deploying a regression model in Amazon SageMaker that predicts housing prices. The model shows high bias (underfitting). Which action is most likely to reduce bias?

Accepted Answer

Add more features or increase model complexity. High bias (underfitting) means the model is too simple to capture the underlying patterns in the data. Adding more features or increasing model complexity (e.g., using polynomial features, deeper trees, or a more flexible algorithm) directly addresses underfitting by giving the model greater capacity to learn from the data. In Amazon SageMaker, this could involve using a more complex built-in algorithm like XGBoost with deeper trees or adding feature engineering transformations in a processing job.

Answer

Reduce the amount of training data

Answer

Increase regularization strength

Answer

Use a simpler model

Question 15

A data scientist is using Amazon SageMaker to train a linear regression model. The training data has 10 features, and the scientist wants to interpret the model's coefficients. Which algorithm should they use?

Accepted Answer

Amazon SageMaker Linear Learner. Amazon SageMaker Linear Learner provides interpretable coefficients, which is essential for understanding the impact of each feature in a linear regression model. Option A is wrong because XGBoost is a tree-based ensemble method that is less interpretable and does not provide linear coefficients. Option B is wrong because K-Means is an unsupervised clustering algorithm, not suited for regression. Option C is wrong because Factorization Machines are designed for high-dimensional sparse data and are not the standard choice for linear regression with 10 features.

Answer

Amazon SageMaker XGBoost

Answer

Amazon SageMaker K-Means

Answer

Amazon SageMaker Factorization Machines

Question 16

A data scientist is performing exploratory data analysis on a dataset with missing values. They want to understand the distribution of each feature and identify outliers. Which AWS service can be used to create visualizations such as histograms and box plots without writing any code?

Accepted Answer

Amazon QuickSight. Amazon QuickSight is a serverless, machine learning-powered business intelligence service that allows users to create interactive dashboards and visualizations without writing code. Option A is wrong because Amazon EMR is a big data platform, not primarily for visualization. Option B is wrong because AWS Glue is used for ETL, not visualization. Option C is correct. Option D is wrong because Amazon SageMaker Studio requires coding for custom visualizations. Option E is wrong because Amazon Athena is a query service, not a visualization tool.

Answer

Amazon EMR

Answer

AWS Glue

Answer

Amazon SageMaker Studio

Answer

Amazon Athena

Question 17

A machine learning engineer is using AWS Step Functions to orchestrate a SageMaker training job followed by a Lambda function for post-processing. The training job completes successfully, but the Lambda function fails with a timeout error. What is the MOST likely cause?

Accepted Answer

The Lambda function execution time exceeds the maximum timeout limit. The Lambda function failed with a timeout error, which directly indicates that its execution duration exceeded the configured maximum timeout limit (default 3 seconds, max 15 minutes). This is the most likely cause because the error message explicitly states 'timeout', and Lambda enforces a hard timeout that terminates the function if it runs longer than the configured limit.

Answer

The Lambda function's IAM role lacks permissions to access the training output

Answer

The Step Functions state machine has a misconfigured retry policy

Answer

The SageMaker training job output data is too large for Lambda to process

Question 18

Which THREE of the following are best practices for feature engineering during EDA? (Select THREE.)

Accepted Answer

Apply log transformation to highly skewed features. Applying a log transformation to highly skewed features helps normalize their distribution, reducing the impact of extreme values and making the data more suitable for many machine learning algorithms that assume normally distributed features. This is a common technique during exploratory data analysis (EDA) to stabilize variance and improve model performance, especially for linear models and neural networks.

Answer

Remove all outliers from the dataset

Answer

Standardize all features to have zero mean and unit variance

Question 19

A team is training a deep learning model on SageMaker using a custom PyTorch container. Training takes 24 hours on a single ml.p3.2xlarge instance. The team wants to reduce training time using distributed training. Which strategy is MOST appropriate?

Accepted Answer

Use data parallelism with Horovod across multiple instances. The training takes 24 hours on a single ml.p3.2xlarge instance with one GPU. To reduce training time, distributing the workload across multiple instances is effective. Data parallelism, where the model is replicated on each instance and each processes different data batches with gradient averaging, is the standard approach for models that fit on a single GPU. Model parallelism is designed for models too large to fit on one GPU, which is not the case here. Therefore, data parallelism with Horovod across multiple instances is the most appropriate strategy.

Answer

Use model parallelism to split the model across multiple GPUs

Answer

Use SageMaker Managed Spot Training to reduce cost

Answer

Use SageMaker Automatic Model Tuning to find optimal hyperparameters

Question 20

A data scientist is training a binary classifier using logistic regression. The dataset has 100,000 samples and 500 features. After training, the model achieves 95% accuracy on the training set but only 70% on the test set. The data scientist suspects overfitting. Which technique would best reduce overfitting while preserving interpretability?

Accepted Answer

Apply L1 regularization (Lasso). L1 regularization (Lasso) adds a penalty equal to the absolute value of the magnitude of coefficients, which drives many feature weights to exactly zero. This performs automatic feature selection, reducing model complexity and overfitting while keeping the model as a simple linear logistic regression, thus preserving interpretability.

Answer

Increase the maximum number of iterations

Answer

Add polynomial features

Answer

Use a random forest model instead