MLS-C01 Practice Test 16 — 15 Questions

Question 1

A media company uses SageMaker to train a neural network for content recommendation. The model uses embeddings for users and items. Training is slow and they want to reduce time. The dataset has 10 million users and 1 million items. They have a cluster of 8 p3.16xlarge instances. Which strategy is most likely to reduce training time?

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

Accepted Answer

Use SageMaker's model parallelism to split the embedding layers across GPUs. SageMaker's model parallelism splits the embedding layers across GPUs, which is essential when the embedding table is too large to fit into the memory of a single GPU. With 10 million users and 1 million items, even with a modest embedding dimension of 256, the embedding layer alone can exceed 10 GB, causing memory bottlenecks that slow training. Model parallelism distributes these large parameters across multiple GPUs, reducing per-GPU memory pressure and enabling larger batch sizes, which directly reduces training time.

Answer

Use data parallelism to replicate the model on each GPU and synchronize gradients

Answer

Reduce the embedding dimension from 256 to 64

Answer

Use a smaller batch size to fit on each GPU

Question 2

A data scientist is using Amazon SageMaker to train a model using a built-in algorithm. The training job fails with an error indicating that the algorithm expects the data to be in recordIO-protobuf format, but the input is CSV. What is the most efficient way to resolve this?

Accepted Answer

Use a SageMaker processing job to convert the CSV data to recordIO-protobuf format.. A SageMaker processing job is the most efficient way to convert CSV to recordIO-protobuf because it runs the conversion at scale within AWS, close to the data in S3, and integrates natively with the SageMaker pipeline. This avoids local bandwidth bottlenecks and keeps the workflow reproducible and automated. The processing job can use a built-in or custom container to perform the conversion and write the output back to S3 for training.

Answer

Change the inference data to recordIO-protobuf format.

Answer

Use a boto3 script to convert the CSV files locally and upload.

Answer

Switch to a different algorithm that accepts CSV format.

Question 3

A data scientist is performing EDA on a dataset with 1,000 features and 10,000 rows. The target variable is binary. After checking for multicollinearity, the scientist finds many pairs of features with correlation > 0.95. Which action should be taken to prepare the data for modeling?

Accepted Answer

For each highly correlated pair, remove one feature based on domain knowledge or higher correlation with target.. When features are highly correlated (e.g., > 0.95), they introduce multicollinearity, which can destabilize coefficient estimates in linear models and reduce interpretability. Removing one feature from each correlated pair based on domain knowledge or its correlation with the target variable preserves predictive power while reducing redundancy. This approach is more targeted than PCA, which transforms features into uncorrelated components but sacrifices interpretability and may not align with the binary target.

Answer

Apply PCA to all features to decorrelate them.

Answer

Standardize all features using StandardScaler.

Answer

Randomly drop half of the correlated features.

Question 4

Which THREE of the following are appropriate methods to reduce overfitting in a decision tree model?

Accepted Answer

Prune the tree after training. Option C (Prune the tree after training) is correct because post-pruning removes branches that add little predictive power, reducing variance and thus overfitting. Option D (Set a minimum number of samples required to split an internal node) is correct because requiring more samples per split prevents the tree from creating splits based on noise in small sample subsets. Option E (Limit the maximum depth of the tree) is correct because capping depth restricts the tree's ability to memorize training data through overly deep, highly specific splits. Option A is incorrect because increasing the number of features considered per split tends to make the tree more expressive and can worsen overfitting. Option B is incorrect because increasing maximum depth allows the tree to grow deeper and fit noise, which increases overfitting.

Answer

Increase the number of features considered for each split

Answer

Increase the maximum depth of the tree

Question 5

A data scientist is trying to create a SageMaker endpoint using an IAM role with the attached policy. The operation fails with 'AccessDenied'. What is the MOST likely cause?

Accepted Answer

The policy does not allow sagemaker:CreateEndpoint.. The error 'AccessDenied' when creating a SageMaker endpoint is most likely due to missing sagemaker:CreateEndpoint permission. To create an endpoint, the IAM role must have the sagemaker:CreateEndpoint action. While InvokeEndpoint is needed for invoking the endpoint after creation, it is not required during creation. S3 bucket access may be needed for model artifacts, but that would typically cause a different error or occur earlier in the process (when creating the model). Therefore, option C is the most likely cause.

Answer

The policy does not allow sagemaker:InvokeEndpoint.

Answer

The policy does not allow access to the S3 bucket for model artifacts.

Answer

The policy does not allow sagemaker:CreateModel.

Question 6

A data scientist is exploring a dataset and wants to check for missing values. Which method is most appropriate to identify the percentage of missing values per column?

Accepted Answer

Use pandas .isnull().sum() in a SageMaker notebook. In a SageMaker notebook, pandas' .isnull().sum() (often divided by len(df) for a percentage) is the standard, direct way to count missing values per column. It operates on the in-memory DataFrame, handles NaN/None uniformly, and gives per-column counts in one line — exactly what the question asks for.

Answer

Use Amazon S3 Select to query missing values

Answer

Use Amazon Athena to run a SELECT COUNT(*) query

Answer

Use Amazon QuickSight to create a missing value dashboard

Answer

Use AWS Glue Crawler to detect missing values

Question 7

A company is performing EDA on a dataset with 10,000 rows and 200 columns. They run a correlation matrix and find many high correlations (|r| > 0.9). What is the best approach to address multicollinearity before modeling?

Accepted Answer

Calculate Variance Inflation Factor (VIF) and remove features with VIF > 10. Variance Inflation Factor (VIF) quantifies how much a feature is correlated with others; removing features with VIF > 10 reduces multicollinearity. Option A is incorrect because standardizing (scaling) does not remove correlations between features. Option C is incorrect because Lasso regression can handle multicollinearity through regularization, but it is a modeling technique and not the best first step during EDA for identifying and removing redundant features. Option D is incorrect because PCA creates orthogonal (uncorrelated) components, which eliminates multicollinearity, but it reduces interpretability and is a dimensionality reduction technique applied during modeling, not an EDA step for directly addressing multicollinearity in the original features.

Answer

Standardize all features

Answer

Use Lasso regression with cross-validation

Answer

Apply Principal Component Analysis (PCA) to all features

Question 8

Refer to the exhibit. A data scientist is trying to run a SageMaker training job using a script that reads training data from 's3://my-bucket/training/data.csv'. The job fails with an access denied error. What is the MOST likely reason?

Accepted Answer

The S3 bucket policy may deny access, or the IAM role lacks necessary permissions beyond GetObject.. The most likely reason for an access denied error when a SageMaker training job reads from S3 is that the SageMaker execution IAM role lacks the necessary S3 permissions, or an explicit bucket policy denies access. SageMaker uses the execution role to access S3, so missing s3:GetObject or a restrictive bucket policy will cause the failure.

Answer

The training job is running in a VPC without S3 VPC endpoint.

Answer

The sagemaker:CreateTrainingJob action is not allowed on the specific resource.

Answer

The S3 path is incorrectly formatted.

Question 9

Refer to the exhibit. What is the recall of the model?

Accepted Answer

0.80. Recall is calculated as True Positives divided by the sum of True Positives and False Negatives. From the confusion matrix, True Positives = 80 and False Negatives = 20, so recall = 80 / (80 + 20) = 0.80. Option B is correct.

Answer

0.85

Answer

0.89

Answer

0.90

Question 10

A machine learning engineer is tuning a gradient boosting model using SageMaker Hyperparameter Tuning. The objective is to minimize MAE. The tuning job uses 20 training jobs. After 10 jobs, the best objective value is 5.2. Which action should the engineer take to potentially improve the result?

Accepted Answer

Increase the total number of training jobs to 50.. Increasing the total number of training jobs from 20 to 50 gives the Bayesian optimization algorithm more opportunities to explore the hyperparameter space and exploit promising regions. With only 10 jobs completed, the tuning job may not have converged to the global minimum of MAE, and additional jobs can refine the search, especially since Bayesian search builds a probabilistic model that improves with more observations.

Answer

Set early stopping to avoid overfitting.

Answer

Change the objective metric to RMSE.

Answer

Switch the tuning strategy from Bayesian to Random search.

Question 11

A company needs to build a data lake on AWS for analytics. The data includes structured, semi-structured, and unstructured data. The solution must support schema-on-read, provide fine-grained access control, and be cost-effective for storing rarely accessed data. Which THREE services should be used? (Choose THREE)

Accepted Answer

AWS Glue Data Catalog for schema-on-read.. Amazon S3 (C) is the correct primary storage layer for a data lake because it natively stores structured, semi-structured, and unstructured data as objects, decoupling storage from compute and enabling schema-on-read through query engines like Athena and Redshift Spectrum. AWS Glue Data Catalog (A) is correct because it provides the centralized metadata repository that defines table schemas and partitions at read time, which is exactly what schema-on-read requires, and it integrates with Lake Formation for fine-grained access control at the table, column, and row level. S3 Lifecycle policies transitioning data to Glacier (E) are correct because they automatically move rarely accessed objects to lower-cost archival storage classes such as Glacier Instant Retrieval, Flexible Retrieval, or Deep Archive, satisfying the cost-effectiveness requirement. Amazon Redshift (B) is not appropriate as the foundational layer here because it is a provisioned data warehouse optimized for structured, schema-on-write workloads rather than a multi-format data lake store. Amazon EMR (D) is a valid processing option but is not required by the stated requirements, since schema-on-read, fine-grained access control, and low-cost archival storage are delivered by the S3, Glue, and lifecycle combination.

Answer

Amazon Redshift for data warehousing.

Answer

Amazon EMR for data processing.

Question 12

Which TWO services can be used to orchestrate a machine learning pipeline?

Accepted Answer

Amazon SageMaker Pipelines. Amazon SageMaker Pipelines (A) is a purpose-built CI/CD service for machine learning that lets you define, automate, and orchestrate end-to-end ML workflows—including data preprocessing, training, tuning, evaluation, and model registration—as a directed acyclic graph (DAG) of steps, so it directly orchestrates an ML pipeline. AWS Step Functions (C) is a general-purpose serverless workflow orchestration service that uses state machines (Amazon States Language) to coordinate AWS services, including SageMaker, Lambda, and Glue, making it a valid choice for orchestrating ML pipelines. The other options do not provide pipeline orchestration: SageMaker Ground Truth (B) is a data-labeling service for building labeled datasets, Amazon Redshift (D) is a cloud data warehouse for analytics, and AWS Glue (E) is a serverless ETL service for data cataloging and transformation—none of these orchestrate the sequence of ML pipeline steps.

Answer

Amazon SageMaker Ground Truth

Answer

Amazon Redshift

Answer

AWS Glue

Question 13

During exploratory data analysis, a machine learning engineer finds that a dataset has a significant number of missing values in a categorical feature with 10 levels. Which approach should they take to handle these missing values before modeling?

Accepted Answer

Create a new category labeled 'Missing' for missing values.. Creating a separate 'Missing' category preserves the missingness pattern and avoids data loss or bias from imputation for categorical features. Option A is incorrect because mean imputation is for numerical features, not categorical. Option C is incorrect because dropping all rows with missing values may discard valuable data and reduce sample size. Option D is incorrect because mode imputation may introduce bias if missingness is not random.

Answer

Impute missing values with the mean of the feature.

Answer

Drop all rows with missing values.

Answer

Impute missing values with the mode of the feature.

Question 14

A data scientist is training a binary classification model on imbalanced data (95% negative, 5% positive). The model achieves 95% accuracy but only 10% recall on the positive class. Which metric should be used to evaluate model performance?

Accepted Answer

F1 score. With 95% negative and 5% positive classes, accuracy is misleading because a model predicting all negatives achieves 95% accuracy but 0% recall on the positive class. The F1 score is the harmonic mean of precision and recall, providing a balanced metric that penalizes models that ignore the minority class. It is the appropriate metric for imbalanced binary classification where both false positives and false negatives matter.

Answer

Accuracy

Answer

Recall

Answer

Precision

Question 15

An organization stores sensitive customer data in S3. A data pipeline uses AWS Glue to transform the data and load it into Amazon Redshift. The security team requires that data be encrypted at rest in S3 and in transit between S3 and Glue, and between Glue and Redshift. Which configuration meets these requirements?

Accepted Answer

Enable S3 default encryption with SSE-KMS, use a VPC endpoint for S3, and configure Glue to use SSL for Redshift connection.. It ensures encryption at rest in S3 via SSE-KMS, encrypts data in transit between S3 and Glue by using a VPC endpoint (which enforces HTTPS/TLS), and encrypts data in transit between Glue and Redshift by configuring SSL for the Redshift connection. SSE-KMS provides envelope encryption with a customer-managed key, while the VPC endpoint and SSL satisfy the in-transit encryption requirements.

Answer

Use S3 client-side encryption, and use VPC Peering between Glue and Redshift.

Answer

Use S3 default encryption with SSE-KMS, and use Network Load Balancer for Redshift.

Answer

Enable S3 server-side encryption with SSE-S3, and use SSL for both Glue connections.