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

 

Question 341

A company wants to predict whether a transaction is fraudulent or legitimate using historical transactions that have already been labeled. Which machine learning task should be selected?

1. Regression
2. Classification
3. Clustering
4. Forecasting

Correct Answer: 2

Explanation

Classification is appropriate when a model must select one category from predefined classes. In this scenario, historical transactions are labeled as fraudulent or legitimate, allowing a supervised learning model to learn patterns associated with each category. The trained model can then classify new transactions. Regression predicts numerical values, clustering discovers groups without predefined labels, and forecasting generally focuses on future values over time. Because the required output is one of two predefined categories based on labeled historical examples, classification is the appropriate machine learning task.

Question 342

A retailer has customer records containing age, location, purchase frequency, and total spending. What are age, location, purchase frequency, and total spending called when they are used as model inputs?

1. Features
2. Labels
3. Predictions
4. Clusters

Correct Answer: 1

Explanation

Features are the input variables that provide information to a machine learning model when it makes predictions. In this example, customer age, location, purchase frequency, and total spending can all serve as features. A label represents the target outcome the model is expected to predict during supervised learning. A prediction is an output produced by the trained model, while a cluster represents a group discovered through an unsupervised learning process. Therefore, the customer attributes used as model inputs are features.

Question 343

A machine learning model is trained to determine whether a customer is likely to cancel a subscription. The training records include the actual cancellation outcome for each customer. What type of learning is being used?

1. Reinforcement learning
2. Unsupervised learning
3. Supervised learning
4. Clustering

Correct Answer: 3

Explanation

Supervised learning uses training data that contains known target outcomes. In this scenario, each customer record includes the actual cancellation result, allowing the model to learn a relationship between customer features and the target outcome. After training, the model can use information from new customers to predict whether they may cancel. Unsupervised learning does not require known target labels, clustering is an unsupervised technique, and reinforcement learning uses feedback or rewards from interactions. Therefore, the presence of known cancellation outcomes indicates supervised learning.

Question 344

A company wants to discover groups of similar products based on sales behavior without defining the groups beforehand. Which technique is most suitable?

1. Regression
2. Classification
3. Clustering
4. OCR

Correct Answer: 3

Explanation

Clustering is an unsupervised machine learning technique that identifies groups of similar observations without requiring predefined categories. A company can provide product information such as sales frequency, average revenue, and purchasing patterns, and the algorithm can discover products with similar characteristics. Classification requires known categories, regression predicts numerical values, and OCR extracts text from images. Because the business does not already know the desired product groups and wants the algorithm to discover them from the available data, clustering is the most suitable technique.

Question 345

A forecasting model predicts the number of website visitors expected each day for the next month. What type of information is central to this workload?

1. Time-series data
2. Image data
3. Audio data
4. Entity data

Correct Answer: 1

Explanation

Time-series data consists of observations associated with specific points or periods in time. Daily website visitor counts form a time-based sequence that can contain trends, recurring patterns, and seasonal behavior. A forecasting model can use historical observations to estimate future visitor numbers. Image and audio data represent different modalities, while entity data is not the primary structure of this forecasting scenario. Since the model uses historical daily measurements to predict future daily values, time-series data is central to the workload.

Question 346

A bank wants to estimate the amount a customer is likely to borrow next year. The output is a monetary value rather than a category. Which task is appropriate?

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

Correct Answer: 2

Explanation

Regression is used to predict numerical values. The amount a customer may borrow next year is a monetary value that can vary across a continuous numerical range, making regression suitable for the task. Features such as previous borrowing behavior, income, account history, and other relevant information could be used to train the model. Classification predicts predefined categories, clustering discovers groups, and entity recognition identifies entities in text. Therefore, because the required output is a numerical loan amount, regression is the appropriate machine learning task.

Question 347

A computer vision application needs to assign a category to an entire image, such as “forest,” “city,” or “beach.” Which capability should be considered?

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

Correct Answer: 2

Explanation

Image classification assigns one or more categories to an image based on its visual content. In this scenario, the application needs to categorize entire images as forest, city, beach, or another supported class. Object detection is different because it identifies individual objects and their locations within an image. Speech recognition converts spoken language into text, while key phrase extraction identifies important concepts in written language. Therefore, image classification is the appropriate computer vision capability when the requirement is to categorize the overall content of an image.

Question 348

A security system needs to identify people, vehicles, and other objects in a camera frame and determine where they appear. Which capability is appropriate?

1. Sentiment analysis
2. Language detection
3. Object detection
4. Text summarization

Correct Answer: 3

Explanation

Object detection identifies objects in images or video and determines their locations, often by providing bounding boxes around detected objects. This makes it suitable for security-camera applications that need to identify people, vehicles, or other objects and determine where they are located within a frame. Sentiment analysis and language detection process language, while text summarization creates shorter versions of written content. Since the system needs both object identification and location information from visual input, object detection is the appropriate capability.

Question 349

A company wants to convert a recorded meeting into searchable text so employees can later find specific statements. Which capability should be used?

1. Speech-to-text
2. Text-to-speech
3. Image classification
4. Object detection

Correct Answer: 1

Explanation

Speech-to-text converts spoken audio into written text. A recorded meeting can therefore be transcribed and stored as searchable text, allowing employees to find statements, topics, or other information within the recording. Text-to-speech performs the opposite operation by generating spoken audio from written content. Image classification and object detection are computer vision capabilities that work with visual information. Because the source is recorded speech and the required output is searchable written content, speech-to-text is the appropriate capability.

Question 350

An AI application needs to identify the language of an incoming written message before sending it to the appropriate translation workflow. Which capability is required?

1. OCR
2. Language detection
3. Object detection
4. Regression

Correct Answer: 2

Explanation

Language detection identifies which language is used in a piece of written text. An application can use this capability to determine whether a message is written in English, Spanish, German, Arabic, or another supported language before selecting an appropriate translation process. OCR extracts text from images, object detection analyzes visual objects, and regression predicts numerical values. Therefore, when the application first needs to determine the language of an incoming written message, language detection is the appropriate capability.

Question 351

A customer sends a long complaint to a support team, and the company wants an AI system to identify the main concepts mentioned in the complaint. Which capability is useful?

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

Correct Answer: 2

Explanation

Key phrase extraction identifies important words and phrases that represent the main concepts in written content. It can help a support organization quickly understand the major topics in a long customer complaint without manually reading every detail. Object detection and image classification work with visual information, while speech synthesis converts text into spoken audio. Key phrase extraction can also support categorization, search, analytics, and downstream language-processing workflows. Therefore, when the requirement is to identify the main concepts within written customer feedback, key phrase extraction is appropriate.

Question 352

A generative AI application should answer questions using information retrieved from an organization’s internal documents. Which concept helps connect the retrieved information with the generated response?

1. Grounding
2. Image segmentation
3. Regression
4. OCR

Correct Answer: 1

Explanation

Grounding provides external information as context for a generative AI model so that its response can be based on relevant source material. In an organizational application, documents can be retrieved from an internal repository and supplied to the model before it generates an answer. This can help make responses more relevant to current company information and reduce reliance on unsupported content. Image segmentation and OCR address visual data, while regression predicts numerical values. Therefore, grounding is the concept that connects retrieved organizational information with a generated response.

Question 353

A generative AI application needs to respond to users in a consistent professional tone. Which approach can help establish the desired behavior?

1. Increasing image resolution
2. Providing clear system instructions
3. Removing all context
4. Using a confusion matrix

Correct Answer: 2

Explanation

Clear system instructions can establish high-level behavioral requirements for a generative AI assistant. Developers can specify that responses should use a professional tone, follow particular formatting rules, or adhere to application-specific constraints. These instructions provide guidance across interactions and can complement user prompts and other application controls. Image resolution concerns visual quality, removing context can reduce the information available to the model, and a confusion matrix evaluates classification results. Therefore, providing clear system instructions is an appropriate way to establish consistent assistant behavior.

Question 354

A company wants to prevent an AI application from returning certain categories of harmful or inappropriate content. Which capability should be considered?

1. Content safety
2. Regression
3. Clustering
4. Time-series forecasting

Correct Answer: 1

Explanation

Content safety capabilities can help organizations identify and manage potentially harmful or inappropriate content in AI applications. Depending on the application, safety mechanisms can evaluate inputs and outputs and apply appropriate filtering, blocking, or review policies. Regression predicts numerical values, clustering groups similar data, and time-series forecasting predicts future values. These machine learning techniques do not directly provide content moderation controls. Therefore, when an AI application needs to manage potentially harmful generated or submitted content, content safety is the relevant capability to consider.

Question 355

An AI team wants to ensure that its evaluation data contains examples representing the different conditions the deployed system is expected to encounter. What quality should the data have?

1. High compression
2. Representativeness
3. Maximum file size
4. Random formatting

Correct Answer: 2

Explanation

Representative evaluation data reflects the types of users, inputs, conditions, and situations that the AI system is expected to encounter after deployment. This allows the team to obtain a more realistic understanding of model performance across relevant scenarios. If important cases are missing from evaluation data, measured performance may not accurately reflect real-world behavior. Compression, file size, and formatting are technical data characteristics but do not describe whether the evaluation dataset adequately represents expected use. Therefore, representativeness is the relevant quality in this scenario.

Question 356

A model’s training performance is poor, and its evaluation performance is also poor. The team suspects that the model cannot capture enough of the underlying patterns. What should they investigate?

1. Overfitting
2. Underfitting
3. Hallucination
4. False positives only

Correct Answer: 2

Explanation

Underfitting occurs when a model is too simple or insufficiently trained to capture important patterns in the data. A common indication is poor performance on both training data and unseen evaluation data. The team may need to investigate model complexity, feature quality, training configuration, or other factors that prevent the model from learning effectively. Overfitting typically produces strong training performance but weaker evaluation performance. Hallucination is associated with unsupported generative AI output, while false positives describe one type of classification error. Therefore, the described pattern suggests underfitting.

Question 357

A deployed model receives data that increasingly differs from the data it was trained on, and its prediction quality begins to decline. What should the team monitor?

1. Data drift
2. Image captioning
3. Speech synthesis
4. OCR

Correct Answer: 1

Explanation

Data drift occurs when the characteristics or distribution of real-world input data change compared with the data used during model development. Such changes can reduce model performance because the model may encounter patterns that differ from those it learned during training. Monitoring input data and model performance can help identify these changes and determine whether retraining or other corrective actions may be necessary. Image captioning, speech synthesis, and OCR are AI capabilities for different workloads. Therefore, data drift should be monitored when changing input patterns are affecting a deployed model.

Question 358

A classification system correctly identifies 850 of 1,000 total cases. Which metric expresses the proportion of all cases that were classified correctly?

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

Correct Answer: 3

Explanation

Accuracy measures the proportion of all predictions that are correct. If a classification system correctly identifies 850 of 1,000 cases, its accuracy is 85 percent. Precision focuses on the proportion of predicted positive cases that are actually positive, while recall measures the proportion of actual positive cases that were successfully identified. Mean absolute error is generally used for numerical prediction problems rather than categorical classification. Therefore, the metric that expresses the proportion of all cases correctly classified is accuracy.

Question 359

A model identifies 80 positive cases correctly out of 100 cases that it predicted as positive. Which metric measures this result?

1. Recall
2. Precision
3. Accuracy
4. F1 score

Correct Answer: 2

Explanation

Precision measures the proportion of predicted positive cases that are actually positive. If the model predicted 100 cases as positive and 80 of those were correct, precision is 80 percent. Recall uses a different denominator: it measures the proportion of all actual positive cases that the model successfully identified. Accuracy considers all correct predictions across both classes, while F1 combines precision and recall. Therefore, because the scenario focuses on the correctness of the model’s positive predictions, precision is the appropriate metric.

Question 360

A model correctly identifies most of the actual positive cases but also produces many incorrect positive predictions. Which trade-off should the team examine?

1. Recall and precision
2. OCR and speech synthesis
3. Regression and clustering
4. Image resolution and file size

Correct Answer: 1

Explanation

Recall and precision describe different aspects of classification performance and can reveal an important trade-off. High recall means the model identifies a large proportion of actual positive cases, while precision measures how many predicted positive cases are actually positive. A model that catches most positives but also generates many incorrect positive predictions may have strong recall but lower precision. The appropriate balance depends on the application’s requirements and the consequences of different errors. Therefore, the team should examine the relationship between recall and precision.