View Full Microsoft AI-900 Exam Dumps and Practice Test Dumps.

 

Question 181

Which machine learning technique is used when a model learns from examples that contain both input features and known target values?

1. Clustering
2. Regression only
3. Supervised learning
4. Unsupervised learning

Correct Answer: 3

Explanation

Supervised learning uses training examples that contain input features together with known target values, or labels. The model learns relationships between the inputs and expected outputs and can then use those relationships to make predictions for new data. Classification and regression are common supervised learning tasks. Clustering is an unsupervised technique, while unsupervised learning generally works without predefined target labels. Because the scenario specifically describes learning from examples where the expected outcomes are already known, supervised learning is the appropriate technique.

Question 182

A retailer wants to divide customers into groups based on purchasing behavior without knowing the groups in advance. Which technique should be considered?

1. Clustering
2. Regression
3. Classification
4. Speech recognition

Correct Answer: 1

Explanation

Clustering is an unsupervised machine learning technique that groups observations according to similarities in their features. A retailer can provide customer information such as purchase frequency, average spending, and product preferences without defining customer groups beforehand. The algorithm can then identify natural groupings in the data. Classification would require predefined categories, while regression predicts numerical values. Speech recognition is unrelated because the data in this scenario represents customer behavior rather than spoken language. Clustering is therefore appropriate for discovering previously unknown customer segments.

Question 183

A model predicts the exact amount a customer is expected to spend next month. Which machine learning task best matches this requirement?

1. Classification
2. Regression
3. Clustering
4. Entity recognition

Correct Answer: 2

Explanation

Regression is used to predict continuous numerical values. In this scenario, the model estimates the amount of money a customer may spend next month, making the target a numerical value rather than a category. Features such as previous spending, purchase frequency, and customer activity could be used as inputs. Classification predicts categories, clustering groups similar observations without predefined labels, and entity recognition identifies entities within text. Since the required output is a numerical spending amount, regression is the appropriate machine learning task.

Question 184

A company wants to determine which words or phrases are most important in a collection of customer comments. Which Azure AI Language capability can help?

1. Speech translation
2. Object detection
3. Key phrase extraction
4. Image classification

Correct Answer: 3

Explanation

Key phrase extraction identifies important words and phrases that represent the main concepts in a piece of text. In customer comments, it can help identify recurring topics such as delivery problems, refunds, product quality, or customer service. This allows organizations to analyze large amounts of feedback more efficiently. Speech translation and image classification address different AI workloads, while object detection is used to identify objects within images. Because the requirement involves identifying important concepts in written comments, key phrase extraction is the appropriate capability.

Question 185

Which Azure AI capability can convert an audio recording of a customer conversation into written text?

1. Speech-to-text
2. Text-to-speech
3. OCR
4. Sentiment analysis

Correct Answer: 1

Explanation

Speech-to-text converts spoken language from audio into written text. This capability can be used for call transcription, meeting notes, voice-controlled applications, and customer-service analysis. Once speech has been converted into text, other language capabilities can process the resulting content, such as sentiment analysis or entity recognition. Text-to-speech performs the opposite operation by converting written text into audio. OCR extracts text from images rather than spoken audio. Therefore, speech-to-text is the appropriate capability for converting a recorded customer conversation into written content.

Question 186

A vision application must identify a vehicle in an image and determine the rectangular region where the vehicle appears. Which capability is required?

1. Image classification
2. Sentiment analysis
3. Object detection
4. Language detection

Correct Answer: 3

Explanation

Object detection identifies objects in images and determines their locations, commonly using bounding boxes or similar regions. A vehicle-detection application can use this capability to recognize a vehicle and indicate where it appears in the image. Image classification can determine categories associated with an image but does not necessarily identify the location of each object. Sentiment analysis and language detection process language rather than visual objects. Because the requirement includes both recognizing the vehicle and locating it, object detection is the appropriate computer vision capability.

Question 187

A company wants an AI assistant to answer questions using a private collection of internal documents. Which architecture can provide those documents as context to a generative AI model?

1. Retrieval-augmented generation
2. Image classification
3. Speech synthesis
4. Regression

Correct Answer: 1

Explanation

Retrieval-augmented generation combines information retrieval with generative AI. When a user asks a question, the application can search a private knowledge source, retrieve relevant documents or passages, and provide that information as context to the language model. The model can then generate a response based on the retrieved material. This approach is useful for enterprise assistants that need to answer questions about internal policies, procedures, products, or documentation. Image classification, speech synthesis, and regression do not provide this document-grounding architecture.

Question 188

A generative AI application should respond with more varied and creative wording. Which adjustment can increase output variability?

1. Lower the temperature
2. Increase the temperature
3. Remove all context
4. Disable the user prompt

Correct Answer: 2

Explanation

Temperature influences the randomness of a generative AI model’s output. Increasing the temperature generally allows the model to produce more varied responses, which can be useful for creative writing or brainstorming scenarios. Lower temperatures generally encourage more predictable and consistent output. The appropriate setting depends on the application requirements because excessive randomness may reduce consistency or factual reliability. Removing context or disabling prompts does not provide meaningful control over generation behavior. Therefore, increasing the temperature can increase variation in generated responses.

Question 189

A classification model produces many false positive results. Which metric is especially useful for determining the proportion of predicted positive cases that are actually positive?

1. Recall
2. Accuracy
3. Precision
4. Mean absolute error

Correct Answer: 3

Explanation

Precision measures the proportion of predicted positive cases that are actually positive. A model producing many false positives may have reduced precision because the predicted-positive group contains a larger number of incorrect results. Recall measures the proportion of actual positive cases that the model successfully identifies, while accuracy measures overall correct predictions. Mean absolute error is primarily used for regression. Examining precision can therefore help an organization understand how reliable its positive predictions are when false positive results are an important concern.

Question 190

Which responsible AI principle is most concerned with ensuring that an AI system operates consistently and safely under expected conditions?

1. Reliability and safety
2. Inclusiveness
3. Transparency
4. Accountability

Correct Answer: 1

Explanation

Reliability and safety focus on ensuring that an AI system performs consistently and safely under the conditions for which it was designed. Organizations should consider testing, monitoring, failure conditions, and appropriate safeguards to reduce the risk of harmful behavior. Inclusiveness focuses on serving diverse users, transparency concerns understanding AI behavior, and accountability concerns responsibility for AI systems and their outcomes. When the primary concern is dependable and safe operation during expected use, reliability and safety are the relevant responsible AI considerations.

Question 191

A document-processing solution handles a common invoice format using an existing model without requiring organization-specific training. What type of model is this likely to be?

1. Custom model
2. Prebuilt model
3. Regression model
4. Clustering model

Correct Answer: 2

Explanation

A prebuilt model is designed for common document-processing scenarios and can often be used without training a model specifically for an organization’s document format. Azure AI Document Intelligence provides prebuilt capabilities for scenarios such as invoices, receipts, and other supported document types. A custom model is more appropriate when specialized documents or fields require organization-specific training. Regression and clustering are machine learning techniques rather than document-processing model categories. Therefore, a commonly supported invoice scenario that does not require custom training is suited to a prebuilt model.

Question 192

An organization wants to know whether its model has become less accurate because the data received after deployment is different from the data used during training. What should the organization monitor?

1. Only the application’s logo
2. Data and model performance
3. The keyboard layout
4. The screen brightness

Correct Answer: 2

Explanation

Monitoring both incoming data and model performance can help identify changes that may affect a deployed AI solution. If production data begins to differ significantly from the training data, the model may no longer perform as expected. Performance monitoring can help identify degradation, while data monitoring can reveal changes in input distributions or other characteristics. Application appearance, keyboard layout, and screen brightness do not provide meaningful information about model behavior. Therefore, monitoring data and model performance is important for maintaining an AI system after deployment.

Question 193

A company wants to identify whether a written message is in English, French, German, or another language before processing it. Which capability should be used?

1. Language detection
2. Entity recognition
3. Key phrase extraction
4. Sentiment analysis

Correct Answer: 1

Explanation

Language detection identifies the language used in a text input. This capability is useful in multilingual applications where messages need to be routed to different processing workflows or translated before further analysis. Entity recognition identifies people, places, organizations, and other entities, while key phrase extraction identifies important concepts. Sentiment analysis evaluates opinions or emotional tone. Because the requirement is specifically to determine which language a written message uses, language detection is the appropriate capability.

Question 194

A company wants to make an AI system accessible and useful to people with different abilities, languages, and backgrounds. Which responsible AI principle is most relevant?

1. Transparency
2. Inclusiveness
3. Privacy
4. Accountability

Correct Answer: 2

Explanation

Inclusiveness focuses on ensuring that AI systems can provide useful experiences for people with diverse abilities, backgrounds, and circumstances. Organizations should consider accessibility and diverse user needs throughout design, development, testing, and deployment. Transparency is concerned with making system behavior understandable, privacy focuses on protecting personal information, and accountability concerns responsibility for AI decisions and outcomes. When an organization wants its AI solution to serve a broad and diverse population effectively, inclusiveness is the responsible AI principle most directly related to that goal.

Question 195

A model performs very well on its training examples but poorly on new examples from the same general problem. What is the most likely issue?

1. Underfitting
2. Overfitting
3. Translation
4. OCR

Correct Answer: 2

Explanation

Overfitting occurs when a model learns the training data too closely and fails to generalize effectively to new examples. Such a model may achieve high performance during training but perform poorly when presented with unseen data. This can happen when the model is overly complex or when the training data is not sufficient to support reliable generalization. Underfitting occurs when a model fails to learn important patterns even from training data. Translation and OCR are AI capabilities rather than explanations for this machine learning behavior.

Question 196

Which type of dataset is normally reserved for evaluating a trained model on examples that were not used during model training?

1. Feature dataset
2. Training dataset
3. Test dataset
4. Label schema

Correct Answer: 3

Explanation

A test dataset contains examples that are kept separate from the training process and can be used to evaluate how well a trained model performs on unseen data. This provides a better indication of generalization than evaluating only on training examples. The training dataset is used to learn model parameters, while features and labels describe information within datasets rather than identifying the final evaluation set. Using a separate test dataset helps reduce the risk of reporting performance that is overly optimistic because the model has already seen the examples.

Question 197

An organization wants to predict next month’s sales using historical monthly sales values. Which type of data is especially relevant to this scenario?

1. Time-series data
2. Image data
3. Audio data
4. Unrelated categorical labels

Correct Answer: 1

Explanation

Time-series data contains observations associated with specific points or periods in time. Historical monthly sales values form a time series because each measurement corresponds to a particular month. An organization can analyze trends, recurring patterns, and other temporal relationships to help estimate future sales. Image and audio data represent different data modalities, while unrelated categorical labels do not capture the temporal sequence needed for this forecasting scenario. Time-series data is therefore particularly relevant when predicting future values based on historical observations ordered over time.

Question 198

A developer needs to make a generative AI assistant follow a specific role and behavioral rules throughout an interaction. Which prompt component is most suitable?

1. Search result
2. System message
3. Test record
4. Image label

Correct Answer: 2

Explanation

A system message provides high-level instructions that establish the behavior and role of a generative AI assistant. It can specify how the assistant should respond, what constraints it should follow, and other application-level requirements. User messages provide individual requests, while search results and test records are sources of information rather than the primary mechanism for defining assistant behavior. Image labels are associated with visual machine learning data. A system message is therefore the appropriate prompt component for establishing consistent behavioral instructions.

Question 199

A model predicts whether an email is spam or legitimate. Which evaluation metric can measure the proportion of all predictions that are correct?

1. Recall
2. Precision
3. Accuracy
4. Mean absolute error

Correct Answer: 3

Explanation

Accuracy measures the proportion of all predictions that are correct. For a spam-classification model, it considers both correctly identified spam and correctly identified legitimate messages along with incorrect predictions. Precision instead focuses on the correctness of predicted positive cases, while recall measures how many actual positive cases were identified. Mean absolute error is commonly used for numerical regression problems. Accuracy can therefore provide an overall measure of classification correctness, although other metrics may also be important when the classes are imbalanced or different errors have different costs.

Question 200

A company wants to protect customer information used by an AI application from unauthorized access and inappropriate use. Which responsible AI concern is most directly involved?

1. Transparency
2. Fairness
3. Privacy
4. Inclusiveness

Correct Answer: 3

Explanation

Privacy focuses on protecting personal and sensitive information throughout the lifecycle of an AI solution. Organizations should consider how customer data is collected, stored, accessed, processed, and shared, while applying appropriate safeguards to reduce unauthorized access or inappropriate use. Transparency concerns making AI behavior understandable, fairness focuses on equitable outcomes, and inclusiveness concerns supporting diverse users. Since the scenario specifically involves protecting customer information from unauthorized access and misuse, privacy is the responsible AI concern most directly associated with the requirement.