MLS-C01 Modeling Practice Test 28 — 15 Questions

Question 1

A data scientist is building a text classification model. The dataset contains 10,000 documents, each labeled with one of 5 categories. Which algorithm is most suitable for this task?

Accepted Answer

Naive Bayes. Naive Bayes is highly suitable for text classification because it models the probability of each category given the document's word features using Bayes' theorem with a strong independence assumption. It performs well on high-dimensional sparse data like bag-of-words or TF-IDF representations, and it is particularly effective when the number of documents (10,000) is moderate relative to the vocabulary size, as it requires relatively little training data to estimate parameters.

Answer

Principal Component Analysis (PCA)

Answer

Linear regression

Answer

k-means clustering

Question 2

A team is training a neural network for image classification using Amazon SageMaker. The training loss decreases rapidly but the validation loss starts increasing after a few epochs. Which action should the team take?

Accepted Answer

Implement early stopping based on validation loss. The described behavior — training loss decreasing while validation loss rises — is the textbook signature of overfitting. Early stopping monitors validation loss during training and halts the job once validation performance stops improving (or begins to degrade), preserving the best model checkpoint. This directly addresses the observed symptom without altering the model architecture or optimization dynamics.

Answer

Reduce the batch size

Answer

Add more convolutional layers

Answer

Increase the learning rate

Question 3

A data scientist is using Amazon SageMaker to train a linear regression model. The target variable is right-skewed. Which transformation should the data scientist apply to the target variable to improve model performance?

Accepted Answer

Log transformation. A right-skewed target variable violates the normality assumption of linear regression and can cause heteroscedasticity and unstable coefficient estimates. Applying a log transformation compresses the long right tail, making the distribution more symmetric and bringing the relationship closer to linear, which improves model fit and prediction accuracy. This is a standard preprocessing step for skewed continuous targets.

Answer

Min-max scaling

Answer

One-hot encoding

Answer

Principal Component Analysis (PCA)

Question 4

A company wants to deploy a machine learning model that predicts customer churn. The model must provide interpretable predictions to explain why a customer is likely to churn. Which algorithm is most appropriate?

Accepted Answer

Decision tree. Decision trees are inherently interpretable because they produce a clear, rule-based structure that shows exactly which features and thresholds lead to a churn prediction. This white-box nature allows stakeholders to trace the reasoning for each prediction, meeting the requirement for interpretability without needing post-hoc explanation methods.

Answer

Gradient boosting machine

Answer

Support vector machine (SVM)

Answer

Deep neural network

Question 5

A data scientist is tuning hyperparameters for an XGBoost model on a large dataset using Amazon SageMaker. The training job is taking too long, and they want to speed up the tuning process. Which strategy is most effective?

Accepted Answer

Use Bayesian optimization. Bayesian optimization is a sequential model-based optimization technique that uses a probabilistic model to guide the search for optimal hyperparameters. It is more sample-efficient than grid or random search, often finding better hyperparameters in fewer iterations, thus speeding up the tuning process. This is especially effective for large datasets where each training run is expensive.

Answer

Use grid search with a fine-grained grid

Answer

Use random search with more iterations

Answer

Reduce the max depth of trees

Question 6

Which TWO of the following are common techniques to handle missing values in a dataset?

Accepted Answer

Remove rows with missing values. Option D (Remove rows with missing values) is correct because listwise deletion, also called complete-case analysis, is a standard and simple technique that drops any record containing a missing value, which is appropriate when missingness is random and the affected rows are few. Option E (Imputation with mean or median) is correct because replacing missing entries with the column's mean (for roughly symmetric numeric data) or median (for skewed data or robustness to outliers) is a widely used single-imputation method that preserves sample size. The other options do not address missing data: A, standardization, rescales features to zero mean and unit variance; B, PCA, is a dimensionality-reduction technique that projects data onto principal components; and C, one-hot encoding, converts categorical variables into binary indicator columns. None of A, B, or C fills in or removes missing values, so they are not missing-value handling techniques.

Answer

Standardization

Answer

Principal Component Analysis (PCA)

Answer

One-hot encoding

Question 7

Which THREE of the following are valid metrics for evaluating a regression model?

Accepted Answer

R-squared (R²). R-squared (R²) (A) is a valid regression metric because it quantifies the proportion of variance in the continuous target explained by the model, with values from -∞ to 1 and 1 indicating perfect fit. Mean Absolute Error (MAE) (C) is valid because it averages the absolute differences between predicted and actual continuous values, giving an interpretable error in the target's units. Root Mean Squared Error (RMSE) (D) is valid because it takes the square root of the mean of squared residuals, penalizing larger errors more heavily and also being expressed in the target's units. F1 score (B) is not a regression metric; it is the harmonic mean of precision and recall used for classification. Accuracy (E) is also a classification metric, measuring the fraction of correct class predictions, and is undefined for continuous regression outputs.

Answer

F1 score

Answer

Accuracy

Question 8

Which TWO of the following are techniques used to reduce overfitting in a neural network?

Accepted Answer

L2 regularization. L2 regularization (C) is correct because it adds a penalty proportional to the sum of squared weights to the loss function, discouraging large weight values and thereby constraining model complexity to reduce overfitting. Dropout (D) is correct because it randomly deactivates a fraction of neurons during each training iteration, forcing the network to learn redundant, robust features and preventing co-adaptation that leads to overfitting. Increasing the number of layers (A) adds parameters and capacity, which typically worsens overfitting rather than reducing it. Batch normalization (B) primarily stabilizes and accelerates training by normalizing layer inputs; while it can have a mild regularizing side effect, it is not a dedicated overfitting-reduction technique. Increasing the learning rate (E) can cause unstable or divergent training and does not address overfitting.

Answer

Increase the number of layers

Answer

Batch normalization

Answer

Increase the learning rate

Question 9

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 10

A company uses SageMaker to train a large language model. The training job is taking too long. The data scientist wants to use distributed training with data parallelism. Which SageMaker feature should be used?

Accepted Answer

SageMaker distributed training libraries. SageMaker distributed training libraries provide data parallelism and model parallelism specifically optimized for AWS infrastructure, enabling efficient multi-GPU/multi-node training of large models. For a large language model with data parallelism, these libraries handle gradient synchronization and communication overhead, directly addressing the slow training job. This is the purpose-built feature for distributed training in SageMaker.

Answer

SageMaker Neo

Answer

SageMaker Processing

Answer

SageMaker Debugger

Question 11

A machine learning engineer is deploying a model to SageMaker for real-time inference. The model is a TensorFlow SavedModel. Which SageMaker capability should be used to create an endpoint?

Accepted Answer

SageMaker hosting with TensorFlow Serving container. SageMaker hosting with TensorFlow Serving container is the correct choice because it provides a managed environment for deploying TensorFlow SavedModels for real-time inference. SageMaker supports custom containers and pre-built containers for popular frameworks like TensorFlow.

Answer

SageMaker Pipelines

Answer

SageMaker Model Monitor

Answer

SageMaker Ground Truth

Question 12

A data scientist is training a deep learning model on a GPU instance. The training loss is decreasing, but the validation loss starts increasing after a few epochs. Which action should the data scientist take to address this?

Accepted Answer

Implement early stopping. The scenario describes overfitting: training loss decreases while validation loss increases. Early stopping is a regularization technique that halts training when validation performance stops improving, preventing overfitting. It is the most direct and appropriate action to address this issue.

Answer

Reduce the batch size

Answer

Increase the learning rate

Answer

Add more layers to the model

Question 13

A data scientist is building a recommendation system for an e-commerce platform. The dataset contains user interactions (clicks, purchases) and item metadata. The scientist wants to use matrix factorization. Which algorithm should be used?

Accepted Answer

SageMaker Factorization Machines. SageMaker Factorization Machines is specifically designed for recommendation systems and matrix factorization tasks. Option A (SageMaker Image Classification) is used for image classification, not matrix factorization. Option B (SageMaker BlazingText) is used for text classification or word embeddings, not for recommendation. Option C (SageMaker XGBoost) is a gradient boosting algorithm for regression and classification, not matrix factorization.

Answer

SageMaker Image Classification

Answer

SageMaker BlazingText

Answer

SageMaker XGBoost

Question 14

A machine learning engineer is deploying a PyTorch model to SageMaker. The model requires custom inference logic. Which approach should the engineer use?

Accepted Answer

Create a custom inference script and use the SageMaker PyTorch container. SageMaker allows you to provide a custom inference script (entry point) when using the PyTorch container, enabling custom inference logic. Option A is wrong because the built-in container as-is would not incorporate custom logic. Option B is wrong because SageMaker Ground Truth is for labeling, not model deployment. Option C is wrong because SageMaker Processing is for data processing, not inference.

Answer

Use a SageMaker built-in PyTorch container as-is

Answer

Use SageMaker Ground Truth to deploy the model

Answer

Use SageMaker Processing to run inference

Question 15

Which THREE techniques can help reduce overfitting in a neural network? (Choose 3)

Accepted Answer

Dropout. Dropout (A) is correct because randomly deactivating a fraction of neurons during each training iteration prevents the network from co-adapting to specific units, effectively acting as an ensemble and reducing overfitting. Early stopping (D) is correct because monitoring validation loss and halting training when it stops improving prevents the model from continuing to fit noise in the training data. L2 regularization (E) is correct because adding a penalty proportional to the sum of squared weights (weight decay) to the loss function constrains weight magnitudes, discouraging overly complex models that memorize training samples. Increasing the number of layers (B) is not correct because adding depth increases model capacity, which typically worsens overfitting rather than reducing it. Using a larger learning rate (C) is not correct because it affects optimization speed and stability, not the model's tendency to overfit, and can even cause divergence.

Answer

Increasing the number of layers

Answer

Using a larger learning rate