Microsoft Azure AI Fundamentals Practice Test (AI-900)

Predicting whether an email is spam or not spam based on its content - This is the correct answer. This is a classification task, where the goal is to predict discrete class labels (spam or not spam) based on input features (the content of the email).

Predicting the future price of a stock based on historical data - This is a regression task, as predicting stock prices involves predicting a continuous numerical value rather than discrete class labels.

Identifying customer clusters based on purchasing behavior - This is a clustering task, which is an unsupervised learning technique that groups customers based on similarities in their purchasing behavior, without predefined class labels.

Estimating the total sales revenue for the next quarter - This is another regression task, as it involves predicting a continuous value (sales revenue) based on input data.

The correct answer is that generative AI models create new, original content by learning patterns from existing data. Generative AI is a category of artificial intelligence models that learn the patterns and structures within training data and then use this knowledge to produce new content, such as text, images, or code .

• Classifying data into predefined categories is a function of classification models, a type of discriminative AI.
• Predicting a continuous numerical value is the function of regression models.
• Analyzing and processing visual information is the primary function of computer vision models.

The Azure AI Translator service specializes in converting text from one language to another, making it ideal for automatically translating product descriptions into multiple languages.

The Azure AI Speech service focuses on speech recognition and synthesis, handling audio inputs and outputs.

The Azure AI Language service deals with understanding and processing natural language but does not provide direct translation capabilities.

Use training data that includes a wide range of demographic groups - This is the correct answer. Ensuring that the training data is diverse and includes a wide range of demographic groups helps reduce biases and ensures that the automated loan approval system treats all applicants fairly. This approach ensures that the model learns from a broad spectrum of data, which can help minimize unfairness in the system.

Exclude any features related to personal characteristics from the data - While this may seem like a good way to prevent bias, excluding personal characteristics (such as age, gender, or race) entirely could lead to a model that lacks important context for making fair and informed decisions.

Increase the complexity of the algorithm to improve accuracy - Increasing algorithm complexity might improve accuracy, but it does not directly address fairness.

Expand the dataset by collecting more data of the same type - Expanding the dataset with more of the same type of data might not necessarily improve fairness if the data itself is biased or unrepresentative.

The design decisions focus on Inclusiveness. Microsoft's Inclusiveness principle states that AI systems should "empower everyone and engage all people, regardless of their backgrounds" and be inclusive "for people of all abilities." Adding screen-reader friendly labels, keyboard navigation, and descriptive alt-text removes barriers for users with disabilities, directly addressing this principle. The other options target different Responsible AI areas: Fairness concerns bias and equal treatment, Transparency is about making the system understandable, and Accountability relates to human oversight.

The Azure AI Vision Service provides capabilities to analyze visual content in images, including object detection, image classification, and optical character recognition (OCR). It is specifically designed for processing and extracting information from images, which aligns with the developer's requirements.

The Azure AI Language Service focuses on natural language processing tasks on text data, not images.

The Azure AI Anomaly Detector Service identifies anomalies in time-series data.

The Azure AI Personalizer Service is used for creating personalized user experiences.

Classification - This is the correct answer. Fraud detection is a classification task because the model needs to classify each transaction as either fraudulent or legitimate, which involves assigning data points to predefined categories or labels.

Regression is used for predicting continuous numerical values for example sales forecasts or prices, not for classifying transactions into categories such as "fraudulent" or "legitimate."

Clustering is an unsupervised learning technique used to group data based on similarities, but it is not suitable for determining whether a transaction is fraudulent or legitimate, which requires labeled data and a classification approach.

Dimensionality Reduction is used to reduce the number of features in the data, typically for improving performance or visualization, but it is not a task in itself for determining fraud or legitimacy.

Deploying the model to a managed batch endpoint is the most appropriate option in this scenario. Managed batch endpoints are designed for processing large volumes of data asynchronously, where predictions do not need to be immediate. This approach is efficient and cost-effective for non-time-sensitive tasks because compute resources are only provisioned for the duration of the job. In contrast, managed online endpoints are for real-time predictions with low latency. Deploying to Azure Kubernetes Service (AKS) or Azure Container Instances (ACI) are options typically associated with online inference for high-scale or dev/test scenarios, respectively, and are not the purpose-built solution for large-scale batch processing in Azure Machine Learning.

Regression is the appropriate technique for predicting continuous numerical values, such as the number of customer service calls. It models the relationship between dependent and independent variables to forecast future values.

Classification is used for predicting categorical outcomes

Clustering groups data points based on similarity without prior labels.

The ability to generate data by learning the data distributions is essential for synthesizing new content. Generative models possess this feature, allowing them to create new data instances similar to the training data.

In contrast, classifying data into predefined classes, predicting continuous values, and identifying clusters involve analyzing existing data without producing new content.

Assigns one or more labels to an entire image based on its content - This is the correct answer. The primary function of an image classification solution is to assign one or more labels to an entire image based on its overall content. For example, it can classify an image as "cat," "dog," or "car" based on what is depicted in the image, without identifying specific object locations within the image.

Detects and localizes individual objects within an image - This describes object detection, which not only identifies objects but also locates them within the image using bounding boxes, whereas image classification only assigns labels to the whole image.

Extracts textual information from images - This describes Optical Character Recognition (OCR), which focuses on extracting and recognizing text from images, not classifying the entire image.

Analyzes facial features to identify individuals - This describes facial recognition, which focuses on identifying and analyzing faces within an image, not general image classification.

Obtaining explicit consent from users and adhering to relevant data protection laws like GDPR is the best approach to ensure compliance when processing personal data. This involves informing users about how their data will be used and ensuring all data handling practices meet legal requirements. While encryption, data anonymization, and restricting data collection are important measures, they alone may not fulfill all legal obligations under privacy laws.

The 'SellingPrice' is the label because it is the output variable that the model aims to predict. In supervised learning, the label is the target variable, and the features are input variables that help predict the label.

'SizeInSquareFeet', 'NumberOfBedrooms', and 'LocationRating' are features used by the model to make predictions.

Key Phrase Extraction is the process of automatically identifying the most significant and relevant phrases within a text. This technique helps in summarizing the main topics or issues discussed, enabling the company to quickly understand customer concerns.

Sentiment analysis identifies the emotional tone behind words, language detection identifies the language of the text, and entity recognition identifies and classifies named entities, such as names of people or organizations, which may not provide the overall themes or issues.

Facial detection solutions are used to locate human faces within digital images or videos, enabling systems to process and analyze facial data. While facial recognition involves identifying and verifying individual identities, facial detection simply identifies the presence and position of faces.

Options related to recognizing identities, translating text, or detecting general objects pertain to other computer vision tasks and not specifically to facial detection.

Entity Recognition identifies and classifies entities like names, organizations, dates and locations in text, converting unstructured data into structured information.

Sentiment Analysis assesses the emotional tone of the text.

**Key Phrase Extraction identifies important terms and phrases.

**Language Modeling predicts word sequences or understands language structure.

Analyzing images and extracting visual information - This is the correct answer. The Azure AI Vision service provides developers with capabilities to analyze images and extract visual information. This includes tasks such as image classification, object detection, and optical character recognition (OCR), helping developers to gain insights from images.

Transcribing spoken language into text - This is the functionality of speech recognition services, not the Azure AI Vision service. Speech recognition is focused on converting spoken language into written text.

Translating text between different languages - This is the functionality of Azure Translator service, which is designed for translating text between languages, not related to image analysis.

Performing sentiment analysis on text data - This is the functionality of text analytics services, which are used for tasks like sentiment analysis, key phrase extraction, and language detection, but it is not related to image analysis.

Sentiment analysis of customer reviews involves processing and understanding human language, which is a key aspect of natural language processing (NLP).

Recognizing objects in images is a computer vision task.

Predicting equipment failures using sensor data is related to predictive analytics.

Translating data into visual charts is data visualization.

Key Phrase Extraction is used to identify the main topics or significant concepts in text data, helping organizations understand commonly discussed themes. This is ideal for summarizing large volumes of feedback to pinpoint areas of interest or concern.

Sentiment Analysis determines the emotional tone behind the text but doesn't highlight specific topics.

Named Entity Recognition identifies and classifies specific entities like names of people, organizations, or locations, but may miss broader topics.

Language Detection determines the language of the text and doesn't assist in extracting thematic content.

Entity Recognition is the appropriate feature for this scenario. It enables the detection and classification of specific entities such as products, brands and locations within text data, allowing the company to extract structured information from unstructured feedback.

Key Phrase Extraction highlights important terms but doesn't categorize them into specific types.

Sentiment Analysis assesses the emotional tone of the text but doesn't identify specific entities.

Language Modeling predicts word sequences and isn't used for extracting or categorizing mentions of specific items.