MLS-C01 Practice Test 48 — 15 Questions

Question 1

A company uses AWS Glue to run ETL jobs that transform data from Amazon S3 to Amazon Redshift. The Glue job writes data to Redshift using the JDBC connection. Recently, the job has been failing with connection timeout errors when writing to Redshift. The Redshift cluster is a 2-node dc2.large cluster. The Glue job processes about 50 GB of data per run. The errors occur sporadically, and the job succeeds after a few retries. The data engineer needs to resolve the issue to prevent job failures. What should the engineer do?

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

Accepted Answer

Modify the Glue job to write data to S3 first, then use the Redshift COPY command to load data.. Writing 50 GB directly to Redshift over JDBC from Glue is slow and prone to connection timeouts because JDBC inserts are row-by-row and the 2-node dc2.large cluster has limited concurrency. The recommended pattern is to have Glue write the transformed data to Amazon S3 in Parquet/CSV, then issue a Redshift COPY command, which uses massively parallel loading and is far more efficient and resilient. This eliminates the sporadic timeouts caused by long-running JDBC writes.

Answer

Increase the Redshift cluster size to 4 nodes.

Answer

Increase the Glue job timeout to 24 hours.

Answer

Use AWS Database Migration Service (DMS) to load data into Redshift.

Question 2

Refer to the exhibit. A data scientist is assigned an IAM policy to deploy a SageMaker model. When the scientist tries to create an endpoint, the action fails with an authorization error. What is the missing permission?

Accepted Answer

iam:PassRole. When creating a SageMaker endpoint, the service must assume an IAM role that has permissions to access required resources (e.g., S3, ECR). The user needs iam:PassRole permission to pass that role to SageMaker. Without it, the CreateEndpoint call fails with an authorization error. The other options are not required for endpoint creation.

Answer

sagemaker:ListEndpoints

Answer

sagemaker:InvokeEndpoint

Answer

sagemaker:UpdateEndpoint

Question 3

A data analyst wants to check for duplicate rows in a dataset stored in S3. Which AWS service can be used to run a SQL query to count duplicates without moving the data?

Accepted Answer

Amazon Athena. Amazon Athena is a serverless interactive query service that allows running standard SQL queries directly on data stored in Amazon S3, without needing to move the data. It can easily count duplicate rows using SQL GROUP BY and HAVING clauses. Option B (Amazon Redshift Spectrum) is wrong because although it can query data in S3, it requires an active Redshift cluster, which is unnecessary for this simple ad-hoc query. Option C (AWS Glue) is wrong because Glue is an ETL service for data preparation and cataloging, not a query engine. Option D (Amazon SageMaker Studio) is wrong because it is an integrated development environment for machine learning, not a SQL query service.

Answer

Amazon Redshift Spectrum

Answer

Amazon SageMaker Studio

Answer

AWS Glue

Question 4

A machine learning engineer is exploring a dataset stored in Amazon S3 that contains 50 million records of customer transactions in Parquet format. The engineer needs to quickly calculate summary statistics (mean, median, standard deviation) for several numerical columns and also visualize the distribution of transaction amounts. The engineer wants to minimize data movement and avoid loading the entire dataset into memory. Which approach is MOST efficient for this exploratory data analysis?

Accepted Answer

Use Amazon Athena to run SQL queries that compute the statistics directly on the Parquet data in S3, and then use Amazon QuickSight to visualize the distribution.. Amazon Athena queries data directly in Amazon S3 without loading it into a separate system, which minimizes data movement and leverages the columnar Parquet format for efficient scans. It supports SQL aggregate functions for summary statistics and integrates with Amazon QuickSight for visualization. This serverless approach avoids cluster management and memory constraints, making it ideal for quick EDA on large datasets.

Answer

Download the Parquet files to a local machine, load them into a pandas DataFrame, and use pandas profiling to compute the statistics and plots.

Answer

Launch an Amazon EMR cluster with Apache Spark, read the Parquet data into a Spark DataFrame, and use Spark SQL to compute the statistics and generate plots.

Answer

Use AWS Glue to crawl the S3 data and create a Data Catalog table, then use Amazon Redshift Spectrum to query the data and compute the statistics.

Question 5

A machine learning team is building a multi-class image classifier using a pre-trained ResNet-50 model in Amazon SageMaker. The dataset has 10 classes but is highly imbalanced, with one class representing 80% of the samples. The team wants to improve model performance on the minority classes. Which TWO of the following approaches are most likely to help? (Select TWO.)

Accepted Answer

Oversample the minority classes in the training data.. Oversampling the minority classes (Option A) directly addresses class imbalance by replicating samples from underrepresented classes, giving the model more exposure to them during training. This is a standard data-level technique that helps the ResNet-50 model learn discriminative features for minority classes without altering the loss function or model architecture.

Answer

Reduce the batch size to increase the frequency of weight updates.

Answer

Increase the number of layers in the model.

Answer

Switch to a focal loss function.

Question 6

Which TWO techniques are used to handle missing values in a dataset before training? (Choose 2.)

Accepted Answer

Mean or median imputation.. Mean or median imputation (A) is a standard missing-value handling technique: it replaces missing entries in a numeric feature with the column's mean (for roughly symmetric distributions) or median (robust to outliers), preserving the row count so the model can still train on that record. Removing rows or columns with missing values (C), often called listwise deletion or column dropping, is another direct missing-data technique that eliminates incomplete data before training, which is valid when the missingness is limited and dropping does not bias the dataset. The other options address different preprocessing goals: min-max scaling (B) rescales numeric features to a fixed range such as [0,1] but does not fill or remove missing entries, one-hot encoding (D) converts categorical variables into binary indicator columns and cannot represent missing values without a separate imputation step, and principal component analysis (E) is a dimensionality-reduction method that generally requires a complete numeric matrix and is not itself a missing-value treatment.

Answer

Min-max scaling.

Answer

One-hot encoding.

Answer

Principal component analysis (PCA).

Question 7

An e-commerce company wants to build a recommendation system. They have user-item interaction data (clicks, purchases) and user demographic data. The goal is to recommend items that a user is likely to purchase. Which approach should be used?

Accepted Answer

Factorization Machines using user-item interactions and user features.. Factorization Machines (FM) can model interactions between users and items while also incorporating user demographic features, making them ideal for this scenario. They capture higher-order interactions and handle sparse data well, which is common in recommendation systems.

Answer

Linear regression on user and item features.

Answer

Collaborative filtering using matrix factorization.

Answer

Content-based filtering using item features.

Question 8

A data scientist is analyzing a large dataset of images stored in Amazon S3. The dataset is used to train a computer vision model. Which THREE EDA steps are appropriate for this image dataset?

Accepted Answer

Compute the distribution of image dimensions (height and width).. Option A is correct because computing the distribution of image height and width is a core EDA step for computer vision datasets, revealing whether images are uniform or require resizing/augmentation before training. Option B is correct because checking for corrupted or unreadable image files (e.g., truncated JPEGs or invalid PNG headers) prevents training failures and silent data loss. Option D is correct because visualizing a sample of images per class verifies label correctness and exposes mislabeled or ambiguous examples that would degrade model performance. Option C is not appropriate because decomposing timestamps for seasonality applies to time series forecasting, not static image training data. Option E is not appropriate because tokenization and stop word removal are NLP text preprocessing techniques, and image filenames carry no meaningful linguistic signal for model training.

Answer

Decompose the time series of image timestamps to detect seasonality.

Answer

Perform tokenization and stop word removal on image filenames.

Question 9

A company runs an e-commerce platform on AWS. They have a SageMaker endpoint serving a product recommendation model. The model uses a custom container with a TensorFlow model. Recently, the endpoint has been returning high latency and occasional 504 errors during peak traffic. The data scientist observes that the model inference time is around 200 ms per request, but the endpoint is configured with a single ml.c5.large instance. The traffic spikes can reach 100 requests per second. The data scientist needs to reduce latency and eliminate 504 errors. Which course of action is most appropriate?

Accepted Answer

Configure the SageMaker endpoint with Application Auto Scaling to scale out based on the 'InvocationsPerInstance' metric, and use a larger instance type such as ml.c5.xlarge. The endpoint is bottlenecked by both instance size and concurrency. With a single ml.c5.large instance handling 100 requests per second and a 200 ms inference time, the instance can only process about 5 requests per second (1000 ms / 200 ms = 5 requests per second per instance). Application Auto Scaling based on the 'InvocationsPerInstance' metric will add instances during traffic spikes, while upgrading to ml.c5.xlarge doubles compute capacity per instance, reducing latency and eliminating 504 errors caused by request queue overflow.

Answer

Use Amazon Elastic Inference to attach an EI accelerator to the endpoint instance

Answer

Switch to a multi-model endpoint to serve multiple models on the same instance

Answer

Replace the SageMaker endpoint with an AWS Lambda function that loads the model from S3 and returns predictions

Question 10

A data scientist is exploring a dataset with a column 'transaction_date'. They want to create features for day of week and month. What is the correct AWS service to schedule a recurring ETL job for this transformation?

Accepted Answer

AWS Glue. AWS Glue is a serverless ETL service that can be scheduled to run recurring jobs for data transformation, such as extracting day of week and month from a date column. Option A is wrong because Amazon Athena is an interactive query service, not an ETL scheduler. Option C is wrong because Amazon SageMaker is a machine learning platform, not an ETL service. Option D is wrong because AWS Lambda can run code on a schedule but is not a full-fledged ETL service for recurring transformations of large datasets.

Answer

Amazon Athena

Answer

Amazon SageMaker

Answer

AWS Lambda

Question 11

A company uses Amazon SageMaker to train a deep learning model using TensorFlow. The training job is failing with an 'OutOfMemory' error. The instance type is ml.p3.2xlarge with 16 GB GPU memory. The model has 10 million parameters. Which THREE actions should be taken to resolve the memory issue? (Choose THREE.)

Accepted Answer

Reduce the batch size. Option A (Reduce the batch size) is correct because the batch size directly determines how many samples are held in GPU memory at once; lowering it reduces activation memory and is the most direct fix for an OutOfMemory error on an ml.p3.2xlarge with 16 GB of GPU memory. Option C (Enable mixed precision training) is correct because using FP16/BF16 instead of FP32 roughly halves activation and gradient memory, and NVIDIA Volta V100 GPUs in the p3 family support Tensor Cores for faster, lower-memory training. Option E (Use gradient accumulation) is correct because it lets you keep an effective large batch size by summing gradients over several small micro-batches, so each forward/backward pass fits in memory while preserving convergence behavior. Option B (Increase the number of epochs) is incorrect because it only lengthens training time and does not reduce per-step memory usage. Option D (Increase the batch size) is incorrect because it would increase GPU memory consumption and worsen the OutOfMemory error.

Answer

Increase the number of epochs

Answer

Increase the batch size

Question 12

A company is training a deep learning model on SageMaker using a large dataset stored in S3. The training job is taking a long time due to I/O bottlenecks. Which action would MOST effectively reduce the I/O bottleneck?

Accepted Answer

Use Pipe input mode in the SageMaker estimator.. SageMaker Pipe input mode streams training data directly from S3 to the training container over a Unix domain socket, avoiding the full dataset download to local EBS storage. This dramatically reduces startup time and I/O bottlenecks for large datasets, making it the most effective single action.

Answer

Use Amazon EFS as the data source.

Answer

Use Amazon FSx for Lustre as the data source.

Answer

Increase the number of training instances.

Question 13

A data scientist is analyzing a dataset and suspects the presence of outliers that could affect the mean and standard deviation. Which TWO methods are robust to outliers for measuring central tendency and dispersion?

Accepted Answer

Interquartile range (IQR). The interquartile range (IQR) is correct because it measures dispersion using the difference between the 75th percentile (Q3) and the 25th percentile (Q1), so extreme values in the tails do not affect it. The median is correct because it is the middle value of an ordered dataset, and its position depends only on rank rather than magnitude, making it resistant to outliers. By contrast, the range is not robust because it is defined as maximum minus minimum, so a single extreme value directly inflates it. Standard deviation is not robust because it squares deviations from the mean, heavily weighting outliers. The mean is not robust because it sums all values and divides by n, so outliers pull it toward themselves.

Answer

Range

Answer

Standard deviation

Answer

Mean

Question 14

A data scientist is training a model using SageMaker's built-in XGBoost algorithm with a large dataset stored in CSV format. The training job is using File mode. The data scientist wants to reduce the time it takes to start training. Which approach would be most effective?

Accepted Answer

Use Pipe mode for the input data channel.. Pipe mode streams data directly from Amazon S3 into the training container, eliminating the need to first download the entire dataset to the EBS volume. This reduces the startup time significantly because training can begin as soon as the first records arrive, rather than waiting for the full download to complete.

Answer

Increase the size of the EBS volume.

Answer

Convert the data to Parquet format.

Answer

Increase the number of training instances.

Question 15

A data scientist is training a random forest model on a dataset with 50 features. After training, the model achieves 98% accuracy on the training set but only 85% on the test set. Which technique is most appropriate to reduce the generalization error?

Accepted Answer

Reduce the maximum depth of each tree. The model shows a large gap between training accuracy (98%) and test accuracy (85%), which is a classic sign of overfitting. Reducing the maximum depth of each tree limits the model's ability to memorize training data, thereby reducing variance and improving generalization. This directly addresses the overfitting problem without discarding useful features or requiring more data.

Answer

Apply Principal Component Analysis (PCA) to reduce dimensionality

Answer

Add more training data

Answer

Increase the number of trees in the forest