AWS Certified Machine Learning
Specialty Practice Exam Questions Well
Detailed Answers Plus Rationales 2026
Q&A Instant Download Pdf
1.
A machine learning team needs to store a large collection of training
datasets in a highly durable, scalable, and cost-effective service. Which
AWS service is most appropriate?
A. Amazon EBS
B. Amazon EFS
C. Amazon S3
D. Amazon FSx
Answer: C. Amazon S3
Amazon S3 provides highly durable object storage and is commonly
used to store machine learning datasets, training artifacts, and model
outputs.
2.
Which AWS service provides managed Jupyter-based environments for
building and training machine learning models?
A. Amazon EC2
B. Amazon SageMaker
C. Amazon Lightsail
D. AWS Lambda
Answer: B. Amazon SageMaker
,SageMaker provides managed environments and tools for preparing
data, developing models, training, tuning, deploying, and monitoring
ML solutions.
3.
A dataset contains a target variable with two possible classes: fraud and
legitimate. What type of machine learning problem is this?
A. Regression
B. Clustering
C. Binary classification
D. Reinforcement learning
Answer: C. Binary classification
Binary classification predicts one of two possible classes.
4.
A model predicts the exact selling price of a house. Which type of
machine learning problem is this?
A. Classification
B. Clustering
C. Regression
D. Association
Answer: C. Regression
Regression predicts a continuous numerical value such as price,
temperature, or revenue.
5.
Which technique is commonly used to reduce the impact of features
that have very different numerical scales?
,A. One-hot encoding
B. Feature scaling
C. Bagging
D. Tokenization
Answer: B. Feature scaling
Scaling transforms numerical features to comparable ranges, which can
improve the performance of many ML algorithms.
6.
A categorical feature contains values such as red, green, and blue.
Which preprocessing method converts these categories into separate
binary features?
A. Normalization
B. Imputation
C. One-hot encoding
D. Standardization
Answer: C. One-hot encoding
One-hot encoding creates a separate binary indicator for each category.
7.
A dataset has missing values in an important numerical feature. Which
approach can replace missing values with the median?
A. Tokenization
B. Median imputation
C. One-hot encoding
D. Bagging
Answer: B. Median imputation
, Median imputation replaces missing numerical observations with the
median value of the available observations.
8.
What is the primary purpose of splitting data into training and test
datasets?
A. Increase storage capacity
B. Reduce AWS costs
C. Evaluate model performance on unseen data
D. Remove all outliers
Answer: C. Evaluate model performance on unseen data
The test set provides an estimate of how well the trained model
generalizes to previously unseen observations.
9.
What is overfitting?
A. A model that is too simple
B. A model that cannot learn from training data
C. A model that learns training data too closely and performs poorly
on unseen data
D. A model with no features
Answer: C. A model that learns training data too closely and performs
poorly on unseen data
Overfitting occurs when a model captures noise or overly specific
patterns in the training data.
10.
Which technique can help reduce overfitting?
Specialty Practice Exam Questions Well
Detailed Answers Plus Rationales 2026
Q&A Instant Download Pdf
1.
A machine learning team needs to store a large collection of training
datasets in a highly durable, scalable, and cost-effective service. Which
AWS service is most appropriate?
A. Amazon EBS
B. Amazon EFS
C. Amazon S3
D. Amazon FSx
Answer: C. Amazon S3
Amazon S3 provides highly durable object storage and is commonly
used to store machine learning datasets, training artifacts, and model
outputs.
2.
Which AWS service provides managed Jupyter-based environments for
building and training machine learning models?
A. Amazon EC2
B. Amazon SageMaker
C. Amazon Lightsail
D. AWS Lambda
Answer: B. Amazon SageMaker
,SageMaker provides managed environments and tools for preparing
data, developing models, training, tuning, deploying, and monitoring
ML solutions.
3.
A dataset contains a target variable with two possible classes: fraud and
legitimate. What type of machine learning problem is this?
A. Regression
B. Clustering
C. Binary classification
D. Reinforcement learning
Answer: C. Binary classification
Binary classification predicts one of two possible classes.
4.
A model predicts the exact selling price of a house. Which type of
machine learning problem is this?
A. Classification
B. Clustering
C. Regression
D. Association
Answer: C. Regression
Regression predicts a continuous numerical value such as price,
temperature, or revenue.
5.
Which technique is commonly used to reduce the impact of features
that have very different numerical scales?
,A. One-hot encoding
B. Feature scaling
C. Bagging
D. Tokenization
Answer: B. Feature scaling
Scaling transforms numerical features to comparable ranges, which can
improve the performance of many ML algorithms.
6.
A categorical feature contains values such as red, green, and blue.
Which preprocessing method converts these categories into separate
binary features?
A. Normalization
B. Imputation
C. One-hot encoding
D. Standardization
Answer: C. One-hot encoding
One-hot encoding creates a separate binary indicator for each category.
7.
A dataset has missing values in an important numerical feature. Which
approach can replace missing values with the median?
A. Tokenization
B. Median imputation
C. One-hot encoding
D. Bagging
Answer: B. Median imputation
, Median imputation replaces missing numerical observations with the
median value of the available observations.
8.
What is the primary purpose of splitting data into training and test
datasets?
A. Increase storage capacity
B. Reduce AWS costs
C. Evaluate model performance on unseen data
D. Remove all outliers
Answer: C. Evaluate model performance on unseen data
The test set provides an estimate of how well the trained model
generalizes to previously unseen observations.
9.
What is overfitting?
A. A model that is too simple
B. A model that cannot learn from training data
C. A model that learns training data too closely and performs poorly
on unseen data
D. A model with no features
Answer: C. A model that learns training data too closely and performs
poorly on unseen data
Overfitting occurs when a model captures noise or overly specific
patterns in the training data.
10.
Which technique can help reduce overfitting?
