IncludeHelp_logo
Skip Navigation Links
×

Trending Technologies MCQs

Home » Trending Technologies MCQs

Automated Machine Learning (AutoML) MCQs (Multiple-Choice Questions)

Practice Automated Machine Learning (AutoML) MCQs to test your understanding of techniques that automate important steps in developing machine learning models. These multiple-choice questions cover automated data preprocessing, feature engineering, algorithm selection, hyperparameter optimization, model evaluation, and deployment. They are useful for students, AI developers, data scientists, machine learning engineers, and candidates preparing for technical interviews. The collection includes foundational and practical questions to strengthen your understanding of AutoML.

Automated Machine Learning (AutoML) MCQs

These Automated Machine Learning multiple-choice questions cover important concepts such as automated model selection, hyperparameter tuning, neural architecture search, Bayesian optimization, grid search, random search, ensemble learning, cross-validation, feature selection, and model deployment. The questions explore how AutoML systems automate machine learning workflows, reduce manual experimentation, and help identify suitable models through conceptual, technical, and scenario-based questions.

These AutoML MCQs help learners understand the benefits, limitations, techniques, tools, and practical considerations involved in automating machine learning workflows. Each question includes an answer and explanation.

List of Automated Machine Learning (AutoML) MCQs

Explore the following 50 MCQs covering AutoML fundamentals, optimization techniques, data preparation, model evaluation, popular tools, applications, challenges, and real-world implementation scenarios.

1. What is Automated Machine Learning (AutoML)?

  1. A technique used exclusively for storing datasets
  2. A process that replaces all data with randomly generated values
  3. An approach that automates selected steps of the machine learning development workflow
  4. A programming language used only for building websites

Answer: C) An approach that automates selected steps of the machine learning development workflow

Explanation:

AutoML automates some or many steps involved in building machine learning models, such as preprocessing, feature selection, algorithm selection, and hyperparameter optimization. The level of automation depends on the system and the workflow being used.

2. What is a primary objective of AutoML?

  1. To reduce manual effort in developing and optimizing machine learning models
  2. To eliminate the need for data quality checks in every project
  3. To guarantee perfect model predictions
  4. To prevent users from evaluating model performance

Answer: A) To reduce manual effort in developing and optimizing machine learning models

Explanation:

AutoML can automate repetitive experimentation and help identify suitable modeling pipelines. It can improve productivity, but users still need to define the problem, provide appropriate data, evaluate results, and consider deployment requirements.

3. Which of the following is a common component of an AutoML workflow?

  1. Manual editing of every prediction
  2. Disabling model evaluation
  3. Removing all input features before training
  4. Automated hyperparameter optimization

Answer: D) Automated hyperparameter optimization

Explanation:

AutoML systems often search for effective hyperparameter configurations using techniques such as random search, Bayesian optimization, or successive halving. Other common components include preprocessing, model selection, and evaluation.

4. What are hyperparameters in machine learning?

  1. Values that are always calculated as model predictions
  2. Configuration settings that control the training process or model structure
  3. Labels assigned to test examples after evaluation
  4. Files containing only the model's output predictions

Answer: B) Configuration settings that control the training process or model structure

Explanation:

Hyperparameters include settings such as learning rate, tree depth, regularization strength, and the number of estimators. They are typically chosen before or during training experiments, unlike model parameters that are learned from training data.

5. What is hyperparameter optimization in AutoML?

  1. The process of searching for hyperparameter settings that improve a chosen evaluation objective
  2. The process of deleting all model parameters
  3. A method for sorting datasets alphabetically
  4. A technique that replaces training with manual predictions

Answer: A) The process of searching for hyperparameter settings that improve a chosen evaluation objective

Explanation:

Hyperparameter optimization evaluates different configurations to find settings that perform well on a defined validation objective. AutoML can automate this search rather than requiring users to test every configuration manually.

6. Which search strategy evaluates every combination in a predefined hyperparameter grid?

  1. Random search
  2. Bayesian optimization
  3. Grid search
  4. Gradient clipping

Answer: C) Grid search

Explanation:

Grid search evaluates combinations from specified sets of hyperparameter values. It is straightforward to implement, but the number of configurations can grow rapidly when many hyperparameters or values are included.

7. How does random search select hyperparameter configurations?

  1. By evaluating every possible configuration without sampling
  2. By sampling configurations from specified search spaces or distributions
  3. By always selecting the first configuration in a list
  4. By choosing values only after examining the final test labels

Answer: B) By sampling configurations from specified search spaces or distributions

Explanation:

Random search samples hyperparameter combinations instead of exhaustively testing every combination. It can explore large search spaces effectively when the evaluation budget is limited, especially when only a subset of hyperparameters strongly affects performance.

8. What is Bayesian optimization in AutoML?

  1. A method that always tests every possible model configuration
  2. A technique for converting numerical features into text
  3. A process that selects models without measuring performance
  4. An optimization approach that uses a probabilistic model of the objective to guide subsequent evaluations

Answer: D) An optimization approach that uses a probabilistic model of the objective to guide subsequent evaluations

Explanation:

Bayesian optimization builds a model of the relationship between hyperparameters and observed performance. An acquisition function helps select promising configurations by balancing exploration of uncertain regions with exploitation of regions expected to perform well.

9. What is the purpose of automated feature selection?

  1. To identify a useful subset of input features for model training
  2. To increase the number of missing values in a dataset
  3. To replace all input variables with the target label
  4. To prevent the model from learning patterns

Answer: A) To identify a useful subset of input features for model training

Explanation:

Feature selection methods identify features that are relevant to the prediction task. Automated selection can reduce dimensionality, training costs, and overfitting risk, although its effectiveness depends on the data and evaluation method.

10. What is automated feature engineering?

  1. A method that removes all original columns from a dataset
  2. A technique that guarantees perfect accuracy without training
  3. The automated creation or transformation of input features to improve model performance
  4. A process that only changes column names

Answer: C) The automated creation or transformation of input features to improve model performance

Explanation:

Automated feature engineering creates useful representations from existing data. Examples include extracting date components, encoding categorical variables, transforming numerical features, and generating interactions between variables.

11. Which of the following is an example of automated data preprocessing?

  1. Manually rewriting every prediction
  2. Automatically imputing missing values and encoding categorical features
  3. Removing the target variable from all evaluation records
  4. Disabling the data validation process

Answer: B) Automatically imputing missing values and encoding categorical features

Explanation:

AutoML pipelines may automatically select or configure preprocessing steps such as missing-value imputation, categorical encoding, and numerical scaling. These transformations should be fitted using training data only to avoid information leakage.

12. What is the purpose of cross-validation in an AutoML workflow?

  1. To make all training examples identical
  2. To eliminate the need for a performance metric
  3. To ensure that the model memorizes the entire dataset
  4. To estimate model performance across multiple training and validation splits

Answer: D) To estimate model performance across multiple training and validation splits

Explanation:

Cross-validation evaluates a model across multiple data splits to obtain a more robust estimate of performance. AutoML systems can use it to compare candidate pipelines, although the appropriate splitting strategy depends on the dataset and problem.

13. Which problem is commonly addressed by AutoML classification workflows?

  1. Predicting discrete categories such as spam or non-spam
  2. Sorting files by their names without learning
  3. Rendering a static HTML page
  4. Compressing images without classifying them

Answer: A) Predicting discrete categories such as spam or non-spam

Explanation:

Classification assigns inputs to discrete classes. AutoML systems can compare candidate classification algorithms, tune their hyperparameters, and select a model using an appropriate metric such as F1-score or accuracy.

14. Which type of problem is typically solved using regression in AutoML?

  1. Assigning a document to a topic label only
  2. Determining whether an email is spam
  3. Predicting a continuous value such as house price
  4. Identifying a file extension from its name

Answer: C) Predicting a continuous value such as house price

Explanation:

Regression predicts numerical values. An AutoML regression workflow may compare algorithms such as linear regression, random forests, and gradient boosting while optimizing hyperparameters against a metric such as mean absolute error.

15. What is Neural Architecture Search (NAS)?

  1. A method for compressing database files
  2. An automated process for searching neural network architectures
  3. A technique for manually labeling all training examples
  4. A process that removes every hidden layer from a neural network

Answer: B) An automated process for searching neural network architectures

Explanation:

Neural Architecture Search explores candidate neural network structures to identify architectures that meet a performance objective. The search may consider layer types, connectivity, width, depth, and computational constraints.

16. Which of the following is a potential benefit of AutoML?

  1. It guarantees that every dataset is unbiased
  2. It removes the need for data governance
  3. It makes every model equally accurate
  4. It can reduce manual experimentation and help users with limited machine learning expertise develop models

Answer: D) It can reduce manual experimentation and help users with limited machine learning expertise develop models

Explanation:

AutoML can make model development more accessible by automating common workflow steps. However, users still need to understand the prediction objective, data quality, evaluation requirements, and limitations of the selected model.

17. What is an ensemble model in AutoML?

  1. A system that combines predictions from multiple models
  2. A model that can only process one input feature
  3. A method for removing all trained models from memory
  4. A technique that guarantees zero prediction errors

Answer: A) A system that combines predictions from multiple models

Explanation:

Ensemble learning combines predictions from multiple models through methods such as voting, averaging, bagging, boosting, or stacking. AutoML systems may build ensembles to improve predictive performance or robustness when the additional complexity is justified.

18. Why might an AutoML system use early stopping?

  1. To ensure every candidate model trains for the same maximum number of epochs regardless of progress
  2. To eliminate the need for a validation dataset
  3. To stop a training process when a monitored condition indicates that continued training is unlikely to be useful
  4. To automatically increase the number of target classes

Answer: C) To stop a training process when a monitored condition indicates that continued training is unlikely to be useful

Explanation:

Early stopping can terminate training when validation performance stops improving or when a resource limit is reached. It can reduce wasted computation and help control overfitting in supported training workflows.

19. What is the role of a search space in AutoML?

  1. It stores only the final test predictions
  2. It defines the candidate algorithms, hyperparameters, or configurations that the system can explore
  3. It determines the physical size of the training computer
  4. It replaces all model evaluation metrics

Answer: B) It defines the candidate algorithms, hyperparameters, or configurations that the system can explore

Explanation:

The search space specifies which model choices and parameter values are available to the AutoML optimizer. A well-designed search space focuses computation on meaningful configurations, while an overly broad space may waste resources.

20. What is the purpose of an objective function in AutoML optimization?

  1. To determine the color of a model visualization
  2. To remove all hyperparameters from the search space
  3. To generate random labels for every example
  4. To define the quantity that the search process aims to optimize, such as validation error or a utility score

Answer: D) To define the quantity that the search process aims to optimize, such as validation error or a utility score

Explanation:

An objective function measures how well a candidate configuration meets a specified goal. AutoML may minimize validation loss or optimize a metric such as F1-score, sometimes subject to latency, memory, or model-size constraints.

21. Which evaluation metric is particularly useful for an imbalanced classification problem?

  1. F1-score
  2. Number of columns in the dataset
  3. Size of the model's filename
  4. Number of training script comments

Answer: A) F1-score

Explanation:

F1-score combines precision and recall using their harmonic mean. It can be more informative than accuracy for imbalanced classification, although the appropriate metric depends on the business objective and the relative costs of different errors.

22. What is data leakage in an AutoML workflow?

  1. A process that makes training data easier to store
  2. A technique for increasing the number of valid features
  3. The unintended use of information during training or model selection that would not be legitimately available at prediction time
  4. A method for automatically correcting every mislabeled example

Answer: C) The unintended use of information during training or model selection that would not be legitimately available at prediction time

Explanation:

Data leakage can occur when preprocessing uses information from validation or test data, or when a feature contains information unavailable at prediction time. Leakage can produce misleadingly high evaluation scores and poor real-world performance.

23. Why should the test dataset remain separate from AutoML model selection?

  1. To ensure the system selects the largest model
  2. To prevent the model from making predictions
  3. To eliminate the need for a validation metric
  4. To provide a more reliable final estimate of performance on unseen data

Answer: D) To provide a more reliable final estimate of performance on unseen data

Explanation:

AutoML may evaluate many candidate pipelines, so repeatedly selecting models based on test results can overfit the model selection process to the test set. A separate test dataset should be reserved for final evaluation after model selection is complete.

24. What is automated model selection?

  1. A process that trains only one fixed algorithm without comparison
  2. A process that compares candidate algorithms or pipelines and selects one based on a defined evaluation objective
  3. A method that assigns random predictions without training
  4. A process that removes all evaluation metrics

Answer: B) A process that compares candidate algorithms or pipelines and selects one based on a defined evaluation objective

Explanation:

Automated model selection allows a system to evaluate different algorithms or complete pipelines and choose a suitable candidate. The choice depends on the specified metric, validation strategy, computational budget, and deployment constraints.

25. Which of the following is a popular open-source library for automated machine learning in Python?

  1. Auto-sklearn
  2. Matplotlib alone as a complete AutoML system
  3. Beautiful Soup
  4. Requests

Answer: A) Auto-sklearn

Explanation:

Auto-sklearn is an AutoML library built around scikit-learn that automates aspects of algorithm selection and hyperparameter optimization. It can also use ensemble methods to combine promising models.

26. What is the primary purpose of TPOT?

  1. To manage relational databases
  2. To generate HTML layouts automatically
  3. To automate the search for machine learning pipelines using optimization techniques, including genetic programming
  4. To compress images without using machine learning

Answer: C) To automate the search for machine learning pipelines using optimization techniques, including genetic programming

Explanation:

TPOT explores machine learning pipelines, including combinations of preprocessing steps and predictive models. Its search can use genetic programming to evolve candidate pipelines according to a chosen evaluation objective.

27. Which platform provides automated machine learning capabilities as part of a cloud-based machine learning service?

  1. A plain text editor without machine learning extensions
  2. Azure Machine Learning
  3. A basic image viewer
  4. A file compression utility

Answer: B) Azure Machine Learning

Explanation:

Azure Machine Learning includes automated machine learning capabilities for supported prediction tasks. It can evaluate candidate pipelines and configurations while allowing users to configure objectives, constraints, and validation settings.

28. What is a key advantage of using cloud-based AutoML services?

  1. They remove all security and privacy responsibilities
  2. They guarantee that every model will be inexpensive to run
  3. They eliminate the need to inspect model outputs
  4. They can provide managed computing resources and integrated tools for model development and deployment

Answer: D) They can provide managed computing resources and integrated tools for model development and deployment

Explanation:

Cloud-based AutoML services may provide scalable compute, experiment tracking, model registration, and deployment integrations. Users must still consider cost, data governance, access control, latency, and the suitability of the generated models.

29. What is the purpose of time-series-aware validation in AutoML?

  1. To preserve temporal ordering and avoid using future observations to predict the past
  2. To randomly mix all time points regardless of the prediction objective
  3. To remove timestamps from every dataset
  4. To guarantee that historical patterns remain unchanged forever

Answer: A) To preserve temporal ordering and avoid using future observations to predict the past

Explanation:

Time-series forecasting often requires chronological validation splits. Randomly mixing past and future observations can create leakage and unrealistic estimates of forecasting performance.

30. Why is feature scaling sometimes included in an AutoML pipeline?

  1. To make every feature contain the same information
  2. To guarantee that all algorithms achieve identical accuracy
  3. To place numerical features on comparable scales when required by the selected algorithms
  4. To replace the target variable with standardized input values

Answer: C) To place numerical features on comparable scales when required by the selected algorithms

Explanation:

Feature scaling can help algorithms that are sensitive to the magnitude of input features, such as many distance-based methods and gradient-based models. Tree-based algorithms often require less scaling, so the preprocessing choice should depend on the pipeline.

31. What is the purpose of automated categorical encoding?

  1. To delete every categorical feature from the dataset
  2. To convert categorical values into representations that a selected model can process
  3. To generate new target labels without evidence
  4. To replace numerical features with file names

Answer: B) To convert categorical values into representations that a selected model can process

Explanation:

Many machine learning algorithms require numerical input. AutoML pipelines may use techniques such as one-hot encoding or other suitable encoders to represent categorical features, with special care for unseen categories and data leakage.

32. What does resource-aware AutoML aim to achieve?

  1. To use the maximum available hardware regardless of the objective
  2. To ignore training time and memory requirements
  3. To select a model solely by its number of parameters
  4. To search for suitable models while respecting constraints such as time, memory, latency, or computational cost

Answer: D) To search for suitable models while respecting constraints such as time, memory, latency, or computational cost

Explanation:

Resource-aware AutoML considers operational limits in addition to predictive quality. For example, a deployment may require low inference latency or a small model footprint, so the highest-accuracy candidate may not be the best overall choice.

33. What is multi-objective optimization in AutoML?

  1. Optimizing two or more objectives, such as predictive performance, latency, and model size
  2. Training a model with no defined objective
  3. Using only one feature for every prediction
  4. Choosing models without measuring their results

Answer: A) Optimizing two or more objectives, such as predictive performance, latency, and model size

Explanation:

Multi-objective AutoML considers multiple goals that may conflict. A smaller model may run faster but produce slightly lower accuracy, so the system may identify a set of trade-off solutions rather than one universally best model.

34. What is a Pareto-optimal solution in multi-objective AutoML?

  1. A solution that is always the worst on every objective
  2. A model that has no trainable parameters
  3. A solution for which no objective can be improved without worsening at least one other objective, among the considered alternatives
  4. A configuration that has not been evaluated

Answer: C) A solution for which no objective can be improved without worsening at least one other objective, among the considered alternatives

Explanation:

A Pareto-optimal solution is not dominated by another candidate that is at least as good on all objectives and better on one or more. AutoML can use Pareto analysis to compare trade-offs between accuracy, latency, cost, and other constraints.

35. Why can automated machine learning be computationally expensive?

  1. It never evaluates candidate models
  2. It may train and evaluate many configurations, pipelines, or architectures
  3. It eliminates all training and validation steps
  4. It always uses a single model configuration

Answer: B) It may train and evaluate many configurations, pipelines, or architectures

Explanation:

AutoML can require substantial computation because many candidates may need to be evaluated. Search budgets, early stopping, parallel evaluation, and resource-aware optimization can help manage the cost.

36. What is the purpose of experiment tracking in an AutoML workflow?

  1. To prevent users from comparing models
  2. To change the test labels automatically
  3. To remove information about model configurations
  4. To record configurations, metrics, artifacts, and results for reproducibility and comparison

Answer: D) To record configurations, metrics, artifacts, and results for reproducibility and comparison

Explanation:

Experiment tracking records information about each model run, including parameters, evaluation results, and saved artifacts. This helps users compare candidate models, reproduce experiments, and identify the configuration selected for deployment.

37. What is model explainability in AutoML?

  1. The ability to understand or interpret factors influencing a model's predictions
  2. The process of increasing a model's file size
  3. A method for hiding every model decision
  4. A technique for replacing all model outputs with random labels

Answer: A) The ability to understand or interpret factors influencing a model's predictions

Explanation:

Model explainability helps users understand how a model makes predictions, for example through feature importance, local explanations, or interpretable model structures. AutoML users may need to consider explainability alongside predictive performance and operational requirements.

38. What is an important limitation of AutoML systems?

  1. They cannot train classification models
  2. They cannot compare model configurations
  3. They may produce poor results when data quality, problem formulation, or evaluation design is inadequate
  4. They always require manual tuning of every parameter

Answer: C) They may produce poor results when data quality, problem formulation, or evaluation design is inadequate

Explanation:

AutoML cannot compensate reliably for every data or problem-definition issue. Biased samples, incorrect labels, leakage, unsuitable metrics, and distribution shifts can all lead to poor model selection or disappointing real-world performance.

39. Why is fairness evaluation important for AutoML models?

  1. To ensure that every group receives identical predictions regardless of the task
  2. To identify potential differences in model performance or harmful outcomes across relevant groups
  3. To eliminate the need for representative data
  4. To ensure that a model always selects the largest architecture

Answer: B) To identify potential differences in model performance or harmful outcomes across relevant groups

Explanation:

Automated model selection may optimize an aggregate metric that hides poor performance for specific groups. Fairness evaluation can reveal disparities and help teams assess whether additional data, different objectives, or other interventions are needed.

40. What is the purpose of model deployment in an AutoML workflow?

  1. To delete the selected model after training
  2. To replace the prediction task with data sorting
  3. To stop the model from accepting new inputs
  4. To make the selected model available for use in an application or production environment

Answer: D) To make the selected model available for use in an application or production environment

Explanation:

Deployment makes a trained model accessible for inference, such as through an API, batch-processing job, or application integration. Production deployment also requires monitoring, version management, security, and maintenance.

41. What is model drift monitoring in a deployed AutoML system?

  1. Monitoring changes in data or model performance that may reduce prediction quality over time
  2. Changing model names at regular intervals
  3. Deleting all incoming data after prediction
  4. Increasing the number of input columns without checking their relevance

Answer: A) Monitoring changes in data or model performance that may reduce prediction quality over time

Explanation:

Production data can change as user behavior, environments, or business conditions evolve. Monitoring input distributions and performance when labels become available can help identify drift and determine whether recalibration, retraining, or a new model is needed.

42. Which statement best describes automated machine learning and human expertise?

  1. AutoML removes the need for humans to define business objectives
  2. Human expertise is never needed when an AutoML system is available
  3. AutoML can automate repetitive modeling tasks, while human judgment remains important for problem formulation, data quality, evaluation, and deployment
  4. AutoML guarantees that no model review is necessary

Answer: C) AutoML can automate repetitive modeling tasks, while human judgment remains important for problem formulation, data quality, evaluation, and deployment

Explanation:

AutoML can reduce manual work, but it does not automatically determine every business requirement or resolve all data issues. Human oversight helps ensure that selected models are appropriate, reliable, fair, and suitable for their intended use.

43. What is automated pipeline optimization?

  1. A method for changing the order of files in a folder
  2. The automated search for effective combinations of preprocessing steps, feature transformations, and model configurations
  3. A process that removes all data preprocessing from a workflow
  4. A technique that evaluates only one fixed pipeline

Answer: B) The automated search for effective combinations of preprocessing steps, feature transformations, and model configurations

Explanation:

Pipeline optimization considers the entire modeling workflow rather than only the final algorithm. AutoML systems can compare different preprocessing, feature selection, and model configurations to identify a suitable end-to-end pipeline.

44. What is the main role of AutoML in predictive maintenance?

  1. To prevent industrial equipment from generating sensor data
  2. To guarantee that machines will never fail
  3. To remove the need for maintenance planning
  4. To automate model development for tasks such as predicting equipment failures from sensor and operational data

Answer: D) To automate model development for tasks such as predicting equipment failures from sensor and operational data

Explanation:

AutoML can compare candidate models for predicting failures, remaining useful life, or maintenance needs. Performance should be evaluated using appropriate temporal splits and metrics that account for the cost of missed failures and false alarms.

45. Why might an AutoML system use stratified cross-validation for classification?

  1. To preserve approximately similar class proportions across folds when appropriate
  2. To guarantee that every fold contains identical records
  3. To remove all minority-class examples from the dataset
  4. To ensure that the training data has no class labels

Answer: A) To preserve approximately similar class proportions across folds when appropriate

Explanation:

Stratified cross-validation attempts to maintain similar class distributions across folds. This can produce more representative evaluations for classification tasks, especially when classes are imbalanced, provided the data structure permits stratification.

46. What is the purpose of model compression in an AutoML deployment workflow?

  1. To increase model size without improving functionality
  2. To remove the need for any performance evaluation
  3. To reduce model size, memory usage, or inference cost while preserving acceptable predictive performance
  4. To guarantee that a model runs on every device without modification

Answer: C) To reduce model size, memory usage, or inference cost while preserving acceptable predictive performance

Explanation:

Model compression techniques such as pruning, quantization, and knowledge distillation can reduce deployment resource requirements. The compressed model should be evaluated to determine whether the performance trade-off is acceptable.

47. Which practice helps make an AutoML experiment reproducible?

  1. Changing the dataset and hyperparameters without recording the changes
  2. Recording data versions, preprocessing steps, random seeds where relevant, configurations, software versions, and evaluation results
  3. Using the test dataset to tune every model repeatedly
  4. Saving only the final accuracy and discarding the model configuration

Answer: B) Recording data versions, preprocessing steps, random seeds where relevant, configurations, software versions, and evaluation results

Explanation:

Reproducibility requires tracking the inputs and settings that affect model training and evaluation. Although some hardware and distributed computations may remain nondeterministic, recording experiment details makes results easier to investigate and reproduce.

48. A retail company wants to predict customer churn from a dataset containing missing values, categorical features, and numerical variables. It wants to compare multiple algorithms without manually tuning each one. Which approach is most appropriate?

  1. Delete all rows containing missing values without examining the impact and select the first model
  2. Train only one algorithm with arbitrary settings and skip validation
  3. Use the test dataset to choose the best hyperparameters
  4. Build an AutoML workflow that handles preprocessing, compares candidate models, tunes hyperparameters, and evaluates performance on appropriate validation data

Answer: D) Build an AutoML workflow that handles preprocessing, compares candidate models, tunes hyperparameters, and evaluates performance on appropriate validation data

Explanation:

An AutoML workflow can automate missing-value imputation, categorical encoding, algorithm selection, and hyperparameter optimization. The company should use a suitable validation strategy, select metrics appropriate for churn prediction, and reserve a separate test set for final evaluation.

49. An AutoML system reports very high validation accuracy for fraud detection, but performance drops sharply after deployment. The team discovers that a feature contains information recorded only after a transaction has been investigated. What is the most likely cause?

  1. Data leakage caused by using information that would not be available at prediction time
  2. Insufficient color contrast in the monitoring dashboard
  3. Too few comments in the model's source code
  4. Excessive use of categorical feature names

Answer: A) Data leakage caused by using information that would not be available at prediction time

Explanation:

The feature includes future information that would not be available when a real-time fraud prediction is made. The team should remove or redesign the feature, reconstruct the evaluation using only information available at prediction time, and validate the revised pipeline on an appropriate holdout dataset.

50. A company needs an AutoML solution for real-time customer risk scoring. It has limited computing resources, an imbalanced dataset, strict latency requirements, and a need to explain important predictions. Which strategy is most appropriate?

  1. Select the most complex model based only on training accuracy and deploy it without testing
  2. Use AutoML to compare candidate pipelines with suitable imbalance-aware metrics, include latency and resource constraints in model selection, evaluate explainability, and validate the selected model on a separate test set
  3. Optimize only for overall accuracy, ignoring minority-class errors and inference latency
  4. Allow the AutoML system to choose a model without defining objectives or reviewing the results

Answer: B) Use AutoML to compare candidate pipelines with suitable imbalance-aware metrics, include latency and resource constraints in model selection, evaluate explainability, and validate the selected model on a separate test set

Explanation:

This scenario requires balancing predictive performance, minority-class detection, inference speed, resource usage, and interpretability. The AutoML workflow should use metrics such as precision, recall, or F1-score as appropriate, evaluate latency and model size, and inspect explanations. A separate test set should be used for final evaluation before deployment.

Comments and Discussions!

Load comments ↻



Top MCQs
Top Programs/Examples
Top Tutorials
About
Student's Section
Subscribe

Copyright © 2026 www.includehelp.com. All rights reserved.