MLS-C01 Practice Test 49 — 15 Questions

Question 1

A data scientist trains a model using SageMaker and notices that the training loss decreases but validation loss increases after a few epochs. What is the MOST likely issue?

Accepted Answer

The model is overfitting the training data.. When training loss continues to decrease while validation loss starts to increase, the model is memorizing the training data rather than generalizing — this is the classic signature of overfitting. The divergence point indicates where the model began fitting noise in the training set at the expense of validation performance.

Answer

The learning rate is too low.

Answer

There is data leakage from validation to training.

Answer

The model is underfitting the training data.

Question 2

A team is building a regression model to predict house prices. The dataset includes a column 'zip_code' with 100 unique values. The data scientist one-hot encodes this column, resulting in 100 new binary columns. The model shows poor performance on a validation set. What is the most likely cause?

Accepted Answer

The model suffers from the curse of dimensionality due to the large number of features.. One-hot encoding 'zip_code' with 100 unique values creates 100 binary features. When combined with other features, the total number of features can be large relative to the number of training samples, leading to the curse of dimensionality. This causes the model to overfit the training data and generalize poorly to the validation set. While multicollinearity among one-hot encoded columns is often low due to their binary nature, the primary issue here is the high dimensionality relative to sample size. Option D correctly identifies this as the most likely cause.

Answer

One-hot encoding introduced multicollinearity among the binary columns.

Answer

One-hot encoding reduced the number of features, causing underfitting.

Answer

The one-hot encoding introduced high variance, but the validation set has low variance.

Question 3

Which TWO options are valid ways to reduce inference latency for a model deployed on a SageMaker real-time endpoint? (Select TWO.)

Accepted Answer

Enable SageMaker Neo to compile the model for the target instance. Option C is correct because SageMaker Neo compiles the model for the specific target instance's hardware and instruction set, producing an optimized executable that runs faster and reduces per-invocation inference latency. Option D is correct because GPU instance types (e.g., ml.p3, ml.g4dn) provide massively parallel compute that accelerates the matrix operations in deep learning models, lowering inference latency compared to CPU-only instances. Option A is wrong because batch transform is designed for offline, large-batch scoring jobs, not low-latency real-time inference. Option B is wrong because adding instances behind a load balancer improves throughput and availability, not the latency of a single inference request. Option E is wrong because the invocation timeout only controls how long the client waits before failing; increasing it does not make the model compute any faster.

Answer

Use SageMaker batch transform instead of real-time endpoint

Answer

Deploy the model to multiple instances behind a load balancer

Answer

Increase the endpoint's invocation timeout

Question 4

Which TWO factors should be considered when choosing between Amazon SageMaker's real-time endpoints and serverless inference? (Select TWO.)

Accepted Answer

GPU requirement. Option A (GPU requirement) is correct because SageMaker serverless inference does not support GPU instances, so any model needing GPU acceleration must use real-time endpoints, which can be backed by GPU-enabled instance types such as ml.g4dn or ml.p3. Option B (Inference traffic pattern) is correct because serverless inference is designed for intermittent or unpredictable traffic and scales to zero when idle, whereas real-time endpoints suit steady, low-latency production traffic and remain provisioned. Option C is not a deciding factor since both endpoint types can be invoked from Lambda via the SageMaker Runtime InvokeEndpoint API. Option D is irrelevant because built-in algorithms are available regardless of the hosting option chosen. Option E is not a primary differentiator, as serverless inference supports models up to a documented size limit but model size alone does not dictate the choice between the two.

Answer

Integration with AWS Lambda

Answer

Availability of built-in algorithms

Answer

Model size in GB

Question 5

A data scientist is training a binary classification model on an imbalanced dataset where the positive class represents 5% of the data. The model achieves 99% accuracy but only identifies 10% of the actual positive cases. Which metric should the data scientist focus on to evaluate the model's performance on the positive class?

Accepted Answer

Recall. Recall (sensitivity) measures the proportion of actual positive cases the model correctly identifies — exactly the metric that is failing here at 10%. With a 5% positive class, accuracy is misleading because predicting 'negative' for everything yields 95% accuracy. Recall directly quantifies how many true positives are being missed.

Answer

Precision

Answer

AUC-ROC

Answer

F1 score

Question 6

A company uses Amazon SageMaker to train a regression model. After training, the data scientist notices that the training loss decreases but validation loss increases after a few epochs. Which EDA technique could have helped predict this behavior?

Accepted Answer

Plot learning curves showing training and validation loss over epochs. Plotting learning curves, which show training and validation loss over epochs, is the correct EDA technique to detect overfitting. The divergence where training loss decreases but validation loss increases is a clear sign of overfitting. Option A (box plots) helps identify outliers but does not directly indicate overfitting. Option C (residual plots) checks for homoscedasticity in regression, not overfitting. Option D (confusion matrix) is used for classification, not regression.

Answer

Create box plots of each feature to identify outliers

Answer

Generate residual plots to check heteroscedasticity

Answer

Plot confusion matrix on the validation set

Question 7

A data scientist needs to version control datasets used for machine learning experiments. Which AWS service should the data scientist use?

Accepted Answer

Amazon S3 with versioning enabled. Amazon S3 with versioning enabled is the correct choice because it provides a simple, scalable, and cost-effective way to version control datasets. S3 versioning preserves every object version, allowing you to retrieve, restore, or compare previous dataset states, which is essential for reproducibility in ML experiments. This directly meets the requirement for dataset version control without additional overhead.

Answer

AWS Lake Formation

Answer

Amazon SageMaker Feature Store

Answer

Amazon SageMaker Model Registry

Question 8

A data engineer needs to schedule an AWS Glue ETL job to run every hour. The job reads from an Amazon DynamoDB table and writes to Amazon S3. Which AWS service should the engineer use to trigger the Glue job on schedule?

Accepted Answer

Amazon EventBridge (CloudWatch Events). Amazon EventBridge (formerly CloudWatch Events) can trigger AWS Glue jobs on a schedule using cron or rate expressions. Option A is incorrect because Amazon Kinesis Data Streams is for real-time streaming data, not scheduling. Option B is incorrect because AWS Step Functions is for orchestrating workflows, but scheduling is typically done via EventBridge. Option D is incorrect because AWS Lambda is a compute service, not a scheduler, though it can be used in conjunction with EventBridge to trigger Glue, but the direct scheduler is EventBridge.

Answer

Amazon Kinesis Data Streams

Answer

AWS Step Functions

Answer

AWS Lambda

Question 9

A model deployed on a SageMaker endpoint is producing predictions that are consistently biased against a certain demographic. Which step should the team take FIRST to address this issue?

Accepted Answer

Use SageMaker Clarify to analyze bias in the training data and predictions. The first step is to analyze the data and model for bias. SageMaker Clarify can detect bias in training data and predictions, making option C the correct first step. Option A (SageMaker Model Monitor) tracks prediction quality but does not specifically analyze bias. Option B (switching algorithm) is a reactive change without understanding the bias source. Option D (retraining with balanced data) is a potential fix but should come after bias analysis.

Answer

Enable SageMaker Model Monitor to track prediction quality

Answer

Switch to a different algorithm that is less prone to bias

Answer

Retrain the model with balanced data

Question 10

A data scientist wants to identify outliers in a dataset. Which TWO techniques are commonly used for outlier detection during EDA?

Accepted Answer

Box plot. A box plot is correct because it visualizes the distribution via quartiles and explicitly flags outliers as points beyond the whiskers, typically those falling below Q1 − 1.5×IQR or above Q3 + 1.5×IQR. Z-score analysis is correct because it standardizes each value as (x − μ)/σ, and observations with |z| greater than a threshold such as 3 are commonly treated as outliers in EDA. A heatmap is not a dedicated outlier-detection technique; it displays a correlation or value matrix and does not flag individual anomalous points. A bar chart is used for comparing categorical frequencies or aggregates and does not reveal statistical outliers. The Pearson correlation coefficient measures the strength and direction of a linear relationship between two variables, not the presence of outliers.

Answer

Heatmap

Answer

Bar chart

Answer

Pearson correlation coefficient

Question 11

A data scientist is performing exploratory data analysis on a dataset stored in Amazon S3 using Amazon SageMaker Studio. The dataset has missing values in several columns. Which approach is the MOST efficient way to handle missing values within SageMaker Studio?

Accepted Answer

Use SageMaker Data Wrangler to impute missing values with mean, median, or mode.. SageMaker Data Wrangler is a purpose-built visual data preparation tool inside SageMaker Studio that provides built-in transforms for imputing missing values using mean, median, or mode, along with other common cleaning operations. It integrates directly with S3 data sources and can export the transformed data or the generated processing code, making it the most efficient approach within Studio.

Answer

Run a Jupyter notebook on a local machine to clean the data and upload back to S3.

Answer

Use AWS Glue to run a find-and-replace operation.

Answer

Write a custom Python script using pandas to drop rows with missing values.

Question 12

A company is using Amazon SageMaker to deploy a model. The model is a large ensemble that requires 8 GB of memory. The company wants to minimize endpoint cost. Which instance type should they choose?

Accepted Answer

ml.m5.large. The ml.m5.large instance provides 8 GB of memory (2 vCPUs, 8 GiB RAM), which meets the model's requirement of 8 GB. It is typically priced lower than ml.c5.xlarge in AWS (e.g., $0.115/hr vs $0.17/hr), making it the most cost-effective choice among instances that can accommodate the memory. The ml.c5.xlarge also provides 8 GB but at a higher cost, while smaller instances lack sufficient memory and larger instances incur unnecessary expense.

Answer

ml.c5.xlarge

Answer

ml.t2.medium

Answer

ml.c5.large

Answer

ml.c5.2xlarge

Question 13

During EDA, a data scientist discovers that two numerical features have a Pearson correlation coefficient of 0.95. Which action should the scientist take to avoid multicollinearity in a linear regression model?

Accepted Answer

Remove one of the features. A Pearson correlation of 0.95 indicates very high linear correlation between the two features, which causes multicollinearity in linear regression — inflating coefficient variance and making estimates unstable. The most direct and standard remediation is to remove one of the correlated features, preserving model interpretability and eliminating the redundancy. While PCA and Ridge can mitigate multicollinearity, they change the model's interpretability or introduce bias, and the question asks for the action to avoid multicollinearity in a linear regression context.

Answer

Apply PCA to the two features

Answer

Use Ridge regression to penalize coefficients

Answer

Create polynomial features from the correlated pair

Answer

Apply min-max scaling to both features

Question 14

A company is migrating its on-premises Apache Hadoop cluster to AWS. The cluster processes large datasets using Spark jobs. The company wants to minimize operational overhead and use native AWS services. Which combination of services should the company use?

Accepted Answer

Amazon EMR with Spark and Amazon S3. Amazon EMR is the native AWS big-data service that runs Apache Spark without managing cluster nodes, and Amazon S3 provides durable, decoupled storage for the datasets. This combination minimizes operational overhead because EMR handles provisioning, scaling, and patching, while S3 decouples compute from storage so clusters can be transient. It is the direct lift-and-shift of a Hadoop/Spark workload to a managed service.

Answer

Amazon Redshift with Spectrum and Amazon S3

Answer

Amazon Athena and AWS Glue

Answer

Amazon EC2 instances with Apache Spark installed and Amazon S3

Question 15

A company is using SageMaker to train a linear learner algorithm. The training log shows that the algorithm converges but the final loss is still high. Which change is most likely to improve the model?

Accepted Answer

Add feature crosses or polynomial features. A high final loss despite convergence indicates that the model is underfitting — the linear decision boundary is too simple to capture the underlying patterns in the data. Adding feature crosses or polynomial features increases the model's expressiveness by introducing non-linear interactions, allowing the linear learner to fit more complex relationships and reduce the loss.

Answer

Reduce the early stopping tolerance

Answer

Increase the maximum runtime

Answer

Increase the number of training instances