Microsoft Azure AI Fundamentals Certification Exam AI-900 Practice Test 2
Q. A company employs a team of customer service agents to provide telephone and email support to customers. The company develops a webchat bot to provide automated answers to common customer queries. Which business benefit should the company expect as a result of creating the webchat bot solution?
increased salesa reduced workload for the customer service agents
improved product reliability
Explanation:
The correct answer is B (a reduced workload for the customer service agents). When a company implements a webchat bot to provide automated answers to common customer queries, the primary and most direct business benefit is reducing the workload on human customer service agents. The bot handles routine, frequently asked questions automatically, allowing agents to focus on more complex issues that require human intervention. While increased sales (A) might be a secondary benefit through improved customer experience, it's not the primary expected outcome. Improved product reliability (C) is unrelated to implementing a customer service chatbot, as the bot doesn't affect the actual products or services offered.
Q. For a machine learning process, how should you split data for training and evaluation?
Use features for training and labels for evaluation.Randomly split the data into rows for training and rows for evaluation.
Use labels for training and features for evaluation.
Randomly split the data into columns for training and columns for evaluation.
Explanation:
The correct answer is B (Randomly split the data into rows for training and rows for evaluation). In machine learning, the standard practice is to split your dataset horizontally (by rows/samples) into separate training and evaluation sets, typically using a random split like 80% for training and 20% for evaluation. Both training and evaluation sets need to contain the same features (input variables) and labels (target variables) - the difference is that the training set is used to teach the model, while the evaluation set is used to test how well the model performs on unseen data. Options A and C are incorrect because both features and labels are needed for both training and evaluation. Option D is incorrect because splitting by columns would separate features from labels, making it impossible to train or evaluate the model properly.
Q. You are developing a model to predict events by using classification. You have a confusion matrix for the model scored on test data as shown in the following exhibit.
Explanation:
The correct answer is 5 correctly predicted positives and 1,033 False negatives. This is based on the confusion matrix shown in the exhibit. In the matrix, the top-left value (5) represents the True Positives (TP), which are instances where the model correctly predicted the positive class. The bottom-left value (1,033) represents the False Negatives (FN), which are actual positives that the model incorrectly predicted as negative. The confusion matrix structure aligns with standard classification evaluation: rows represent predicted values and columns represent actual values. Therefore, the first row (predicted positive) and first column (actual positive) intersection gives TP = 5, and the second row (predicted negative) and first column (actual positive) gives FN = 1,033. This interpretation is further supported by the definitions provided in the reference, which clarifies TP and FN in terms of classifier performance.
Q. You build a machine learning model by using the automated machine learning user interface (UI). You need to ensure that the model meets the Microsoft transparency principle for responsible AI. What should you do?
Set Validation type to Auto.Enable Explain best model.
Set Primary metric to accuracy.
Set Max concurrent iterations to 0.
Explanation:
The correct answer is B (Enable Explain best model). Microsoft's transparency principle for responsible AI emphasizes that AI systems should be understandable and interpretable to users and stakeholders. By enabling 'Explain best model' in the automated ML interface, you generate model explanations that help users understand how the model makes predictions, which features are most important, and why certain decisions are made. This directly supports transparency by making the model's behavior interpretable. The other options do not relate to transparency: setting validation type to Auto (A) is about model validation methodology, setting primary metric to accuracy (C) is about performance optimization, and setting max concurrent iterations to 0 (D) would prevent the model from running at all.
Q.
Explanation:
The correct answers are: 'No' for forecasting housing prices, 'Yes' for identifying suspicious sign-ins, and 'No' for predicting diabetes. This is because anomaly detection refers to identifying patterns in data that do not conform to expected behavior. Identifying suspicious sign-ins by detecting deviations from usual login patterns is a textbook example of anomaly detection, as it focuses on detecting unusual or potentially fraudulent activity. In contrast, forecasting housing prices is a regression task where the model predicts a continuous variable based on historical trends this is not about spotting anomalies. Similarly, predicting whether a patient will develop diabetes based on medical history is a classification or prediction task aimed at determining future outcomes rather than identifying outliers or unusual cases. The key distinction lies in the intent: anomaly detection is about finding rare or unusual events, not predicting typical or expected outcomes.
Q.
Explanation:
The correct answer is 'reliability and safety.' This is because the handling of unusual or missing values directly impacts how reliably and safely an AI system performs. Microsoftâs principle of 'reliability and safety' for responsible AI emphasizes that systems must be resilient, robust, and able to handle edge cases without failure. If an AI system cannot properly manage unexpected inputs like missing or abnormal values, its predictions and decisions could become unreliable or unsafe. Therefore, managing such data scenarios is essential to maintain system trustworthiness and to prevent unintended behaviors in real-world conditions, aligning directly with this responsible AI principle.
Q.
Explanation:
The correct answer is 'reliability and safety.' This is because the handling of unusual or missing values directly impacts how reliably and safely an AI system performs. Microsoftâs principle of 'reliability and safety' for responsible AI emphasizes that systems must be resilient, robust, and able to handle edge cases without failure. If an AI system cannot properly manage unexpected inputs like missing or abnormal values, its predictions and decisions could become unreliable or unsafe. Therefore, managing such data scenarios is essential to maintain system trustworthiness and to prevent unintended behaviors in real-world conditions, aligning directly with this responsible AI principle.
Q. You are designing an AI system that empowers everyone, including people who have hearing, visual, and other impairments. This is an example of which Microsoft guiding principle for responsible AI?
fairnessinclusiveness
reliability and safety
accountability
Explanation:
The correct answer is B (inclusiveness). Microsoft's inclusiveness principle for responsible AI focuses on designing AI systems that empower everyone and engage people across all abilities, backgrounds, and demographics. When you design an AI system specifically to accommodate people with hearing, visual, and other impairments, you are directly implementing the inclusiveness principle by ensuring the technology is accessible to users with diverse needs and capabilities. While fairness (A) deals with treating all people equitably, reliability and safety (C) focuses on consistent and secure performance, and accountability (D) involves taking responsibility for AI systems, inclusiveness specifically addresses designing for accessibility and universal usability across all user groups.
Q.
Explanation:
The correct matches are: 'Reliability and safety' with ensuring AI systems operate as designed and handle unexpected conditions; 'Accountability' with implementing processes to allow human override of AI decisions; and 'Privacy and security' with providing consumers control over data use and storage. These align with Microsoftâs responsible AI principles. 'Reliability and safety' ensures systems can function under varied conditions without malfunction or being manipulated. 'Accountability' emphasizes human oversight and responsibility in AI decision-making, ensuring humans retain control. 'Privacy and security' focuses on protecting personal data and giving users transparency and control over how their information is handled by AI systems.
Q.
Explanation:
The correct answer is 'reliability and safety.' This is because developing an AI system for self-driving cars requires it to function correctly even in unexpected conditions. The Microsoft principle of 'reliability and safety' for responsible AI emphasizes the importance of systems performing as intended, being resilient, and safely handling unforeseen scenarios. Self-driving cars operate in dynamic environments where they must adapt to rapidly changing situations. Ensuring consistent and safe operation under such conditions directly aligns with this principle, making it essential for building trust in autonomous vehicle technologies.
Q. You are building an AI system. Which task should you include to ensure that the service meets the Microsoft transparency principle for responsible AI?
Ensure that all visuals have an associated text that can be read by a screen reader.Enable autoscaling to ensure that a service scales based on demand.
Provide documentation to help developers debug code.
Ensure that a training dataset is representative of the population.
Explanation:
The correct answer is C (Provide documentation to help developers debug code). Microsoft's transparency principle for responsible AI requires that AI systems be understandable and interpretable to users, developers, and stakeholders. Providing comprehensive documentation that helps developers understand how the system works, how to debug issues, and how decisions are made directly supports transparency by making the AI system's behavior and functionality clear and accessible. Option A relates to inclusiveness (accessibility), option B relates to reliability and safety (system performance), and option D relates to fairness (avoiding bias). Only option C directly addresses making the AI system transparent and understandable to those who work with it.
Q.
Explanation:
The correct answer is 'reliability and safety.' This is because developing an AI system for self-driving cars requires it to function correctly even in unexpected conditions. The Microsoft principle of 'reliability and safety' for responsible AI emphasizes the importance of systems performing as intended, being resilient, and safely handling unforeseen scenarios. Self-driving cars operate in dynamic environments where they must adapt to rapidly changing situations. Ensuring consistent and safe operation under such conditions directly aligns with this principle, making it essential for building trust in autonomous vehicle technologies.
Q. You are developing a natural language processing solution in Azure. The solution will analyze customer reviews and determine how positive or negative each review is. This is an example of which type of natural language processing workload?
sentiment analysislanguage detection
entity recognition
key phrase extraction
Explanation:
The correct answer is A (sentiment analysis). Sentiment analysis is a natural language processing technique specifically designed to determine the emotional tone or attitude expressed in text, categorizing it as positive, negative, or neutral. Since the solution described analyzes customer reviews to determine how positive or negative each review is, this is a classic example of sentiment analysis. Language detection (B) identifies what language text is written in, entity recognition (C) identifies and extracts specific entities like names, places, or organizations from text, and key phrase extraction (D) identifies the main topics or important phrases in text. None of these other options focus on determining the emotional polarity of the content.
Q. What are three Microsoft guiding principles for responsible AI? Each correct answer presents a complete solution. NOTE: Each correct selection is worth one point.
knowledgeabilitydecisiveness
inclusiveness
fairness
opinionatedness
reliability and safety
Explanation:
The correct answers are C (inclusiveness), D (fairness), and F (reliability and safety). Microsoft has six guiding principles for responsible AI: Fairness (ensuring AI systems treat all people fairly), Reliability and Safety (ensuring AI systems perform reliably and safely), Privacy and Security (protecting user data and privacy), Inclusiveness (ensuring AI systems empower everyone and engage people), Transparency (ensuring AI systems are understandable), and Accountability (ensuring people are accountable for AI systems). The question asks for three principles, and from the given options, inclusiveness, fairness, and reliability and safety are the three that match Microsoft's official responsible AI principles. Knowledgeability, decisiveness, and opinionatedness are not among Microsoft's established responsible AI principles.
Q. To complete the sentence, select the appropriate option in the answer area. Returning a bounding box that indicates the location of a vehicle in an image is an example of
image classificationobject detection
optical character recognizer (OCR)
semantic segmentation
Explanation:
The correct answer is B (object detection). Object detection is a computer vision technique that not only identifies what objects are present in an image but also locates them by drawing bounding boxes around them. When a system returns a bounding box that indicates the location of a vehicle in an image, it is performing object detection - identifying that there is a vehicle (classification) and providing its precise location (localization) with coordinates. Image classification (A) only identifies what's in an image without location information, OCR (C) is specifically for recognizing text characters in images, and semantic segmentation (D) involves pixel-level classification of every part of an image rather than bounding box detection.
Q. To complete the sentence, select the appropriate option in the answer area. _______ is used to generate additional features.
Feature engineeringFeature selection
Model evaluation
Model training
Explanation:
The correct answer is A (Feature engineering). Feature engineering is the process of creating new features or transforming existing features from raw data to improve machine learning model performance. This includes generating additional features through techniques like combining existing features, creating polynomial features, binning continuous variables, or extracting new information from existing data. Feature selection (B) is about choosing the most relevant features from existing ones, not generating new ones. Model evaluation (C) is about assessing model performance, and model training (D) is about teaching the algorithm using the prepared features. Only feature engineering specifically involves the creation and generation of additional features.
Q. You run a charity event that involves posting photos of people wearing sunglasses on Twitter. You need to ensure that you only retweet photos that meet the following requirements: â Include one or more faces. â Contain at least one person wearing sunglasses. What should you use to analyze the images?
the Verify operation in the Face servicethe Detect operation in the Face service
the Describe Image operation in the Computer Vision service
the Analyze Image operation in the Computer Vision service
Explanation:
The correct answer is B (the Detect operation in the Face service). The Face service's Detect operation can identify faces in images and return detailed face attributes, including accessories like sunglasses. This operation can detect multiple faces and provide information about whether each face is wearing sunglasses, which perfectly matches both requirements: detecting one or more faces and identifying if at least one person is wearing sunglasses. The Verify operation (A) is used to compare faces for identity verification, not for detecting face attributes. The Describe Image operation (C) provides general descriptions of images but doesn't give specific details about face accessories. The Analyze Image operation (D) can detect objects and people but doesn't provide the specific face attribute details needed to determine if someone is wearing sunglasses.
Q. When you design an AI system to assess whether loans should be approved, the factors used to make the decision should be explainable. This is an example of which Microsoft guiding principle for responsible AI?
transparencyinclusiveness
fairness
privacy and security
Explanation:
The correct answer is A (transparency). Microsoft's transparency principle for responsible AI requires that AI systems be understandable and interpretable to users and stakeholders. When designing a loan approval system where the decision-making factors should be explainable, you are implementing transparency by ensuring that the AI system's reasoning and decision process can be understood and explained to applicants, regulators, and other stakeholders. This allows people to understand how and why decisions are made. While fairness (C) is also important in loan decisions to avoid discrimination, the specific requirement that factors be 'explainable' directly relates to transparency. Inclusiveness (B) focuses on accessibility for all users, and privacy and security (D) deals with protecting user data.
Q.
Explanation:
The correct answers are: Statement 1 - Yes, Statement 2 - No, Statement 3 - No. Statement 1 is correct because providing explanations for credit loan application outcomes exemplifies the Microsoft transparency principle, which emphasizes making AI decision-making processes clear and understandable to users. Statement 2 is incorrect because a triage bot prioritizing insurance claims based on injuries actually demonstrates the fairness principle rather than reliability and safety, as it focuses on equitable treatment and unbiased resource allocation. Statement 3 is incorrect because offering AI solutions at different prices across sales territories does not represent the inclusiveness principle, which is about ensuring AI systems are accessible to people with diverse abilities and backgrounds, not about geographic pricing strategies.
Q.
Explanation:
The correct answers are: Statement 1 - Yes, Statement 2 - No, Statement 3 - No. Statement 1 is correct because providing explanations for credit loan application outcomes exemplifies the Microsoft transparency principle, which emphasizes making AI decision-making processes clear and understandable to users. Statement 2 is incorrect because a triage bot prioritizing insurance claims based on injuries actually demonstrates the fairness principle rather than reliability and safety, as it focuses on equitable treatment and unbiased resource allocation. Statement 3 is incorrect because offering AI solutions at different prices across sales territories does not represent the inclusiveness principle, which is about ensuring AI systems are accessible to people with diverse abilities and backgrounds, not about geographic pricing strategies.
Q.
Explanation:
The correct sequence for deploying an Azure Machine Learning model as a service is: 1) Data preparation, 2) Model training, 3) Model evaluation. This follows the standard machine learning workflow where you first prepare and clean your data to ensure it's suitable for training. Next, you train the model using the prepared data to learn patterns and relationships. Finally, you evaluate the model's performance using metrics and validation techniques to ensure it meets the required standards before deployment. Data encryption and model retraining are not part of the initial deployment sequence - encryption is typically handled at the infrastructure level, and retraining occurs after the model is already deployed and being monitored in production.
Q. You are building an AI-based app. You need to ensure that the app uses the principles for responsible AI. Which two principles should you follow? Each correct answer presents part of the solution. NOTE: Each correct selection is worth one point.
Implement an Agile software development methodologyImplement a process of AI model validation as part of the software review process
Establish a risk governance committee that includes members of the legal team, members of the risk management team, and a privacy officer
Prevent the disclosure of the use of AI-based algorithms for automated decision making
Explanation:
The correct answers are B (Implement a process of AI model validation as part of the software review process) and C (Establish a risk governance committee that includes members of the legal team, members of the risk management team, and a privacy officer). Option B supports the reliability and safety principle by ensuring AI models are properly validated before deployment, while option C supports the accountability principle by establishing proper governance and oversight structures. Option A (Agile methodology) is a general software development practice not specifically related to responsible AI principles. Option D contradicts the transparency principle, as responsible AI requires disclosure and explainability of AI system usage, not hiding it from users.
Q. To complete the sentence, select the appropriate option in the answer area. According to Microsoft's _______ principle of responsible AI, AI systems should NOT reflect biases from the data sets that are used to train the systems.
accountabilityfairness
inclusiveness
transparency
Explanation:
The correct answer is B (fairness). Microsoft's fairness principle for responsible AI specifically addresses the need to ensure that AI systems treat all people fairly and do not reflect or amplify biases present in training data. This principle focuses on preventing discrimination and ensuring equitable treatment across different groups of people. When AI systems reflect biases from training data, they can perpetuate unfair treatment or discrimination against certain groups. Accountability (A) is about taking responsibility for AI systems, inclusiveness (C) is about ensuring AI empowers everyone regardless of ability, and transparency (D) is about making AI systems understandable - none of these directly address the issue of bias in training data like the fairness principle does.
Q.
Explanation:
The correct matches are: Feature engineering matches with 'Featurizing the values of a confusion matrix' (Model evaluation), Feature selection matches with 'Splitting a dataset into months, days, and year bands' (Feature engineering), and Model deployment matches with 'Tracking temperature and pressure to train a weather model' (Feature selection). Feature engineering involves creating new features from existing data, such as extracting time components from dates or transforming raw data into more useful formats. Feature selection is the process of choosing the most relevant features from the available dataset to improve model performance and reduce complexity. Model deployment refers to putting a trained model into production where it can make predictions on new data, such as using sensor data to monitor environmental conditions in real-time applications.
Q. To complete the sentence, select the appropriate option in the answer area. Data values that influence the prediction of a model are called
dependent variablesfeatures
identifiers
labels
Explanation:
The correct answer is B (features). In machine learning, features are the input variables or data values that influence and are used by a model to make predictions. Features are the independent variables that the model analyzes to determine patterns and relationships. Dependent variables (A) refer to the output or target variable that the model is trying to predict, not the input values that influence the prediction. Labels (D) are the correct answers or target values used during training (similar to dependent variables). Identifiers (C) are unique values used to distinguish records but typically don't influence predictions. Features are specifically the data values that serve as inputs to influence the model's predictions.
Q. You have the Predicted vs. True chart shown in the following exhibit. Which type of model is the chart used to evaluate?
classificationregression
clustering
Explanation:
The correct answer is B (regression). A Predicted vs. True chart (also known as a scatter plot of predicted vs actual values) is specifically used to evaluate regression models. This type of chart plots the predicted continuous values against the actual continuous values, allowing you to assess how well the model's predictions align with the True values. The ideal result would show points forming a straight diagonal line, indicating perfect predictions. The chart shows continuous numerical values on both axes, which is characteristic of regression problems. Classification models (A) typically use confusion matrices, ROC curves, or precision-recall curves for evaluation, not predicted vs. True scatter plots. Clustering models (C) use different evaluation methods like silhouette scores or within-cluster sum of squares, as they don't have 'True' labels to compare against.
Q. Which type of machine learning should you use to predict the number of gift cards that will be sold next month?
classificationregression
clustering
Explanation:
The correct answer is B (regression). Regression is used when you need to predict a continuous numerical value, such as the number of gift cards that will be sold next month. Since you're predicting a specific quantity (a number), this is a regression problem. The model will output a continuous value representing the predicted count of gift cards. Classification (A) is used when you need to predict categories or classes (like 'high sales', 'medium sales', 'low sales'), not specific numerical values. Clustering (C) is an unsupervised learning technique used to group similar data points together, not to make predictions about future values. Regression is specifically designed for predicting numerical outcomes like sales quantities, prices, temperatures, or counts.
Q. You have a dataset that contains information about taxi journeys that occurred during a given period. You need to train a model to predict the fare of a taxi journey. What should you use as a feature?
the number of taxi journeys in the datasetthe trip distance of individual taxi journeys
the fare of individual taxi journeys
the trip ID of individual taxi journeys
Explanation:
The correct answer is B (the trip distance of individual taxi journeys). Features are the input variables that influence the prediction, and trip distance is a logical feature that would strongly correlate with taxi fare - longer trips typically cost more. Option A (number of taxi journeys in the dataset) is a dataset-level statistic, not a feature of individual journeys. Option C (the fare of individual taxi journeys) is actually the target variable (label) that you're trying to predict, not a feature. Option D (trip ID) is just an identifier that doesn't provide meaningful information to predict fare. Good features for predicting taxi fare would include trip distance, trip duration, time of day, pickup/dropoff locations, and other factors that logically influence the cost of a taxi ride.
Q. You need to predict the sea level in meters for the next 10 years. Which type of machine learning should you use?
classificationregression
clustering
Explanation:
The correct answer is B (regression). Regression is used when you need to predict continuous numerical values, and sea level in meters is a continuous measurement. Since you're predicting specific numerical values (sea level measurements) over time, this is a classic regression problem. The model will output continuous values representing the predicted sea level for each time period. Classification (A) is used for predicting categories or discrete classes, not continuous measurements. Clustering (C) is an unsupervised learning technique used to group similar data points, not to make predictions about future values. Time series regression or forecasting models would be particularly appropriate for this type of temporal prediction problem.
Q.
Explanation:
The correct answers are: Statement 1 - Yes, Statement 2 - No, Statement 3 - Yes, Statement 4 - No. Statement 1 is True because Automated Machine Learning (AutoML) is specifically designed to automate the time-consuming and iterative tasks involved in machine learning model development, such as feature engineering, algorithm selection, and hyperparameter tuning. Statement 2 is False because AutoML cannot automatically infer training data from use cases - you must still provide the training dataset; AutoML automates the model building process, not data collection. Statement 3 is True because AutoML systems automatically run multiple training iterations with different algorithms and hyperparameters, then score and rank them based on specified metrics to find the best performing model. Statement 4 is False because while AutoML can automatically select appropriate algorithms and tune parameters, you still need to specify the dataset and define what you want to predict (the target variable); AutoML cannot automatically understand or infer the prediction task without human input.
Q. To complete the sentence, select the appropriate option in the answer area. A banking system that predicts whether a loan will be repaid is an example of the _______ type of machine learning.
classificationregression
clustering
Explanation:
The correct answer is A (classification). Classification is used when you need to predict discrete categories or classes. In this case, the banking system is predicting whether a loan will be repaid or not - this is a binary classification problem with two possible outcomes: 'repaid' or 'not repaid' (or 'approved' or 'denied'). The model assigns each loan application to one of these discrete categories based on the input features. Regression (B) is used for predicting continuous numerical values, such as the exact amount that might be repaid or the probability as a continuous score. Clustering (C) is an unsupervised learning technique used to group similar data points together, not to make predictions about specific outcomes.
Q.
Explanation:
The correct answers are: Statement 1 - Yes, Statement 2 - No, Statement 3 - No. Statement 1 is True because labeling (or data labeling) is indeed the process of tagging training data with known correct values or annotations to indicate what the machine learning model should predict - this creates the ground truth for supervised learning. Statement 2 is False because you should never evaluate a model using the same data used to train it, as this leads to overfitting and overly optimistic performance estimates; instead, you should use separate validation or test datasets to get accurate performance measurements. Statement 3 is False because accuracy is not always the primary metric for measuring model performance - depending on the problem type and business requirements, other metrics like precision, recall, F1-score, AUC-ROC, or domain-specific metrics may be more appropriate, especially in cases with imbalanced datasets or when different types of errors have varying costs.
Q. Which service should you use to extract text, key/value pairs, and table data automatically from scanned documents?
Form RecognizerText Analytics
Language Understanding
Custom Vision
Explanation:
The correct answer is A (Form Recognizer). Form Recognizer is specifically designed to extract text, key/value pairs, and table data from scanned documents and forms. It uses OCR (Optical Character Recognition) combined with machine learning to understand document structure and extract structured information from various document types including invoices, receipts, forms, and business cards. Text Analytics (B) analyzes text that has already been extracted for sentiment, key phrases, and entities, but doesn't extract text from scanned documents. Language Understanding (C) is used for natural language processing to understand user intents and entities in conversational applications. Custom Vision (D) is used for image classification and object detection, not for document text extraction.
Q. To complete the sentence, select the appropriate option in the answer area. The ability to extract subtotals and totals from a receipt is a capability of the _______ service.
Custom VisionForm Recognizer
Ink Recognizer
Text Analytics
Explanation:
The correct answer is B (Form Recognizer). Form Recognizer is specifically designed to extract structured information from documents, including receipts. It can identify and extract key financial data such as subtotals, totals, taxes, merchant information, and line items from receipts and invoices. The service uses pre-built models for common document types like receipts that can automatically recognize and extract these specific fields. Custom Vision (A) is used for image classification and object detection, not document data extraction. Ink Recognizer (C) is for recognizing handwritten content from digital ink, not for extracting structured data from receipts. Text Analytics (D) analyzes already-extracted text for insights like sentiment and key phrases, but doesn't extract text from document images.
Q. You use Azure Machine Learning designer to publish an inference pipeline. Which two parameters should you use to consume the pipeline? Each correct answer presents part of the solution. NOTE: Each correct selection is worth one point.
the model namethe training endpoint
the authentication key
the REST endpoint
Explanation:
The correct answers are C (the authentication key) and D (the REST endpoint). When you publish an inference pipeline in Azure Machine Learning designer, it creates a REST web service that can be consumed by client applications. To consume this published pipeline, you need two essential parameters: the REST endpoint (D) which is the URL where you send your inference requests, and the authentication key (C) which is required to authenticate and authorize access to the web service. The model name (A) is not a parameter needed to consume the pipeline - it's embedded within the published service. The training endpoint (B) is not relevant for consuming an inference pipeline, as inference pipelines are used for making predictions, not for training models.
Q. To complete the sentence, select the appropriate option in the answer area. From Azure Machine Learning designer, to deploy a real-time inference pipeline as a service for others to consume, you must deploy the model to
a local web serviceAzure Container Instances
Azure Kubernetes Service (AKS)
Azure Machine Learning compute
Explanation:
The correct answers are B (Azure Container Instances) and C (Azure Kubernetes Service (AKS)). When deploying a real-time inference pipeline from Azure Machine Learning designer as a service for others to consume, you need to deploy to a compute target that can host web services. Azure Container Instances (ACI) is suitable for low-scale, CPU-based workloads and testing scenarios, while Azure Kubernetes Service (AKS) is used for high-scale, production deployments that may require GPU support and auto-scaling. A local web service (A) is only for local testing and development, not for others to consume. Azure Machine Learning compute (D) is used for training workloads and batch inference, not for real-time inference services that need to be accessible via REST endpoints.
Q. To complete the sentence, select the appropriate option in the answer area. Predicting how many hours of overtime a delivery person will work based on the number of order received is an example of
classificationclustering
regression
Explanation:
The correct answer is C (regression). Regression is used when predicting continuous numerical values, and 'hours of overtime' is a continuous numerical measurement. The model is trying to predict a specific number of hours based on the input feature (number of orders received), which is a classic regression problem. The output could be any continuous value like 2.5 hours, 4.75 hours, etc. Classification (A) would be used if you were predicting categories like 'low overtime', 'medium overtime', or 'high overtime' rather than the specific number of hours. Clustering (B) is an unsupervised learning technique used to group similar data points together, not to make predictions about specific outcomes.
Q.
Explanation:
The correct answers are: Statement 1 - Yes, Statement 2 - Yes, Statement 3 - No. Statement 1 is True because Azure Machine Learning designer does provide a drag-and-drop visual interface that allows users to build, test, and deploy machine learning models without writing code by connecting pre-built modules on a visual canvas. Statement 2 is True because the designer enables you to save your machine learning workflows as pipeline drafts, allowing you to preserve your progress, make iterative improvements, and reuse components across different projects. Statement 3 is False because Azure Machine Learning designer does not support custom JavaScript functions - it primarily uses Python-based modules and pre-built components; while you can create custom modules, they would typically be written in Python rather than JavaScript, as the Azure ML platform is primarily Python-based for custom scripting and module development.
Q.
Explanation:
The correct answers are: Household Income should be classified as 'A feature' and House Price Category should be classified as 'A label'. In machine learning terminology, features (also called input variables or predictors) are the data points used to make predictions, while labels (also called target variables or dependent variables) are what you're trying to predict. Since the goal is to train a model that predicts house price categories based on household income and postal code, the Household Income serves as input data (a feature) that helps the model learn patterns, while House Price Category is the output we want to predict (the label). The model will use the relationship between household income, postal code, and historical house price categories to make future predictions about house price categories for new data.
Q. To complete the sentence, select the appropriate option in the answer area. Azure Machine Learning designer lets you create machine learning models by
adding and connecting modules on a visual canvasautomatically performing common data preparation tasks
automatically selecting an algorithm to build the most accurate model
using a code-first notebook experience
Explanation:
The correct answer is A (adding and connecting modules on a visual canvas). Azure Machine Learning designer is a drag-and-drop visual interface that allows users to create machine learning models by connecting pre-built modules on a visual canvas without writing code. While the designer can help with data preparation tasks (B), it doesn't automatically perform them - users still need to select and configure the appropriate modules. The designer doesn't automatically select algorithms (C) - users choose which algorithm modules to use. Option D describes Azure Machine Learning notebooks/compute instances, not the designer, which is specifically designed as a no-code/low-code visual tool.