MLS-C01 Modeling Practice Test 35 — 15 Questions

Question 1

A data scientist is building a recommender system using collaborative filtering. The dataset is sparse (99% missing values). Which algorithm is best suited?

Accepted Answer

Matrix Factorization (e.g., SVD). Matrix factorization (e.g., SVD) is best suited for sparse collaborative filtering because it learns latent factors that capture underlying user-item interactions, effectively handling the 99% missing values by generalizing patterns rather than relying on explicit pairwise similarities. Unlike memory-based methods, it decomposes the sparse user-item matrix into lower-dimensional representations, enabling accurate predictions even when most entries are unobserved.

Answer

Random Forest

Answer

K-Nearest Neighbors

Answer

Hidden Markov Model

Question 2

A data scientist wants to automate the selection of optimal hyperparameters for a model. Which SageMaker feature should be used?

Accepted Answer

SageMaker Automatic Model Tuning. SageMaker Automatic Model Tuning (AMT) is the correct feature because it automates hyperparameter optimization by running multiple training jobs with different hyperparameter combinations, using algorithms like Bayesian optimization or random search to find the best set. This directly addresses the requirement to automate selection of optimal hyperparameters.

Answer

SageMaker Debugger

Answer

SageMaker Model Monitor

Answer

SageMaker Experiments

Question 3

Which TWO of the following are valid techniques to handle missing data in a dataset?

Accepted Answer

Mean imputation. Mean imputation (C) is a valid missing-data handling technique because it replaces each missing value in a numeric column with the mean of the observed values in that column, preserving the dataset size and avoiding loss of records. Deleting rows with missing values (E), often called listwise deletion, is also a standard valid technique because it removes incomplete records so that analyses run only on fully observed cases. Normalizing the data (A) is a scaling technique that rescales numeric features (for example to 0–1 or z-scores) and does not address missing values. Adding a constant value of 0 (B) is not a recognized missing-data imputation method; it would distort distributions and is not equivalent to principled zero imputation. SMOTE (D) is an oversampling technique for class imbalance that generates synthetic minority-class samples, not a method for handling missing data.

Answer

Normalizing the data

Answer

Adding a constant value of 0

Answer

Synthetic Minority Over-sampling (SMOTE)

Question 4

Which THREE of the following are common causes of overfitting in machine learning models?

Accepted Answer

Using a complex model like a deep neural network on a small dataset. Option A is correct because a highly complex model such as a deep neural network has enormous capacity and can memorize the idiosyncrasies and noise of a small dataset instead of learning generalizable patterns, which is a classic overfitting cause. Option B is correct because when the number of trainable parameters greatly exceeds the number of training samples, the model has enough degrees of freedom to fit the training data almost perfectly, including its noise, leading to poor generalization. Option D is correct because training for too many epochs lets the model continue optimizing the training loss past the point of best generalization, so it starts fitting noise and the validation/test error rises even as training error keeps falling. Option C is not a cause of overfitting; a large dataset with many samples generally reduces overfitting by providing more evidence and constraining the model. Option E is not a cause of overfitting either; regularization techniques such as L1/L2 penalties, dropout, or early stopping are explicitly used to prevent or reduce overfitting.

Answer

Having a large dataset with many samples

Answer

Using regularization techniques

Question 5

Which TWO of the following are examples of unsupervised learning tasks?

Accepted Answer

Dimensionality reduction using PCA. Dimensionality reduction using PCA (B) is an unsupervised learning task because PCA finds principal components that capture variance in the data without using any labeled target values. Clustering customer segments (D) is also unsupervised, as algorithms like k-means or DBSCAN group similar data points based on feature similarity without predefined labels. In contrast, classifying emails as spam or not spam (A) and sentiment analysis of product reviews (C) are supervised classification tasks that require labeled training data. Predicting house prices (E) is supervised regression, since it learns a mapping from features to a continuous labeled target.

Answer

Classifying emails as spam or not spam

Answer

Sentiment analysis of product reviews

Answer

Predicting house prices

Question 6

A data scientist is training a binary classification model on an imbalanced dataset where the positive class is rare. The model currently achieves 95% accuracy but only 10% recall on the positive class. Which metric should the data scientist prioritize to improve model performance?

Accepted Answer

F1 score. The F1 score is the harmonic mean of precision and recall, making it the best single metric to optimize when the positive class is rare and both false positives and false negatives are costly. With 95% accuracy but only 10% recall, the model is likely predicting the majority class almost exclusively, so improving recall without sacrificing precision is critical — the F1 score directly balances this trade-off.

Answer

AUC-ROC

Answer

Precision

Answer

Accuracy

Question 7

A machine learning engineer is deploying a model to Amazon SageMaker for real-time inference. The model requires low latency and must handle variable traffic patterns. Which SageMaker feature should the engineer use to automatically scale the number of instances based on demand?

Accepted Answer

SageMaker automatic scaling. SageMaker automatic scaling (Application Auto Scaling) is the correct feature because it allows the engineer to define scaling policies (e.g., based on CPU utilization or request latency) that automatically adjust the number of instances behind a SageMaker endpoint in response to real-time traffic patterns. This ensures low latency by maintaining sufficient capacity during spikes and reducing costs during lulls, without manual intervention.

Answer

Amazon EC2 Auto Scaling

Answer

Elastic Inference

Answer

SageMaker Batch Transform

Question 8

A data scientist is training a deep learning model for image classification using Amazon SageMaker. The training job is taking too long. The data scientist wants to speed up training by using distributed training across multiple GPUs. Which SageMaker feature or configuration should the data scientist use?

Accepted Answer

SageMaker Data Parallelism library. The SageMaker Data Parallelism library is specifically designed to distribute training across multiple GPUs by splitting the input data across workers, which reduces per-GPU computation time and accelerates training for deep learning models. This library uses optimized all-reduce algorithms (e.g., Ring AllReduce) to synchronize gradients efficiently, making it ideal for speeding up image classification tasks that are data-intensive.

Answer

SageMaker Debugger

Answer

Model parallelism in SageMaker

Answer

SageMaker hyperparameter tuning

Question 9

A company is building a recommendation system for an e-commerce platform. The data includes user-item interactions and features such as user demographics and item categories. Which algorithm would be most appropriate for generating personalized recommendations?

Accepted Answer

Factorization Machines. Factorization Machines (FM) are specifically designed for recommendation tasks with sparse, high-dimensional data like user-item interactions. They model pairwise feature interactions (e.g., user demographics × item categories) using factorized parameters, enabling personalized recommendations even when many user-item pairs are unobserved. This makes FM far more effective than tree-based or clustering methods for collaborative filtering and feature-rich recommendation scenarios.

Answer

XGBoost

Answer

k-means clustering

Answer

Principal Component Analysis (PCA)

Question 10

A data scientist notices that a linear regression model trained on a dataset has high variance. The model performs well on the training data but poorly on the test data. Which action is most likely to reduce the variance?

Accepted Answer

Apply L2 regularization to the model. High variance indicates the model is overfitting to the training data. L2 regularization (ridge regression) adds a penalty proportional to the square of the magnitude of the coefficients, which shrinks them toward zero. This reduces the model's sensitivity to noise in the training data, thereby lowering variance and improving generalization to the test set.

Answer

Decrease the amount of training data

Answer

Increase the number of gradient descent iterations

Answer

Add more features to the model

Question 11

A machine learning team is using Amazon SageMaker to train a model. The training data is stored in an S3 bucket. The team wants to ensure that the training job can access the data securely without using long-lived AWS credentials. Which approach should the team use?

Accepted Answer

Specify an IAM role in the SageMaker training job configuration. SageMaker training jobs can assume an IAM role specified in the job configuration to obtain temporary security credentials via AWS Security Token Service (STS). This allows the training job to access the S3 bucket securely without embedding long-lived AWS access keys in code or configuration files.

Answer

Store AWS access keys in the training script

Answer

Use an S3 bucket policy that allows public access

Answer

Create a new IAM user for each training job

Question 12

A company's machine learning model is overfitting to the training data. The data scientist has already tried reducing the model complexity and adding regularization, but the model still overfits. Which technique could the data scientist use to further reduce overfitting?

Accepted Answer

Use data augmentation to increase the training dataset size. Data augmentation artificially increases the size and diversity of the training dataset by applying transformations (e.g., rotations, flips, noise injection) to existing samples. This exposes the model to more varied examples, reducing its tendency to memorize noise and improving generalization — directly countering overfitting when other methods have failed.

Answer

Decrease the batch size

Answer

Increase the number of training epochs

Answer

Increase the learning rate

Question 13

A data scientist is training a neural network using a custom loss function. The training process converges, but the model's performance on the validation set is poor. The data scientist suspects that the model is overfitting. Which action should the data scientist take to diagnose overfitting?

Accepted Answer

Plot the training and validation loss over epochs. Plotting the training and validation loss over epochs is the standard diagnostic technique for detecting overfitting. If the training loss continues to decrease while the validation loss plateaus or increases, it indicates that the model is memorizing the training data rather than generalizing. This visual comparison directly confirms overfitting, allowing the data scientist to take corrective action such as regularization or early stopping.

Answer

Add more layers to the network

Answer

Increase the learning rate

Answer

Compute the confusion matrix on the training set

Question 14

Which THREE of the following are valid approaches for deploying a machine learning model to an Amazon SageMaker endpoint for real-time inference?

Accepted Answer

Use a SageMaker Inference Pipeline with multiple containers. Option A is correct because a SageMaker Inference Pipeline lets you chain multiple containers (for example, a preprocessing container plus a model container) behind a single real-time endpoint, which is a supported deployment approach for real-time inference. Option B is correct because SageMaker provides pre-built containers for built-in algorithms, and you can deploy those directly to a real-time endpoint without building custom images. Option E is correct because SageMaker supports bringing your own Docker container that conforms to the SageMaker inference container contract, allowing custom models to be hosted on a real-time endpoint. Option C is not correct because Amazon EMR is a big data processing service for running frameworks like Spark and Hadoop, not a service for hosting SageMaker real-time inference endpoints. Option D is not correct because AWS Lambda runs function code in response to events and is not the mechanism for deploying a model to a SageMaker endpoint, even though Lambda can invoke a SageMaker endpoint.

Answer

Use Amazon EMR to host the model

Answer

Deploy the model as an AWS Lambda function

Question 15

A company is building a binary classifier to predict equipment failure. The dataset has 99% negative (no failure) and 1% positive (failure) examples. The data scientist uses a random forest model with default settings. The model achieves 99% accuracy on the test set but fails to identify any actual failures. Which metric should the data scientist use to evaluate the model?

Accepted Answer

Recall. Recall (sensitivity) measures the proportion of actual positive cases correctly identified. With 99% negative examples, a model can achieve 99% accuracy by simply predicting 'no failure' for all instances, but this yields 0% recall for the failure class. Since the goal is to detect rare failures, recall is the appropriate metric to evaluate the model's ability to find positive cases.

Answer

RMSE

Answer

R-squared

Answer

Precision