Describe Core Machine Learning Concepts AI-900 Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A media company wants to build a model that recommends articles to users. They have a dataset of user interactions: user ID, article ID, time spent, and a 'liked' indicator. The team plans to use supervised learning. Which column should be the label, and why?

   • AThe 'liked' indicator, because it represents the outcome the model should predict
   • BArticle ID, because the model must predict which article is recommended
   • CTime spent, because it measures engagement
   • DUser ID, because the model must learn user preferences
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A logistics company wants to predict delivery delays. They have a dataset with columns: OrderID, Distance, WeatherCondition, TrafficLevel, and Delayed (Yes/No). They plan to train a supervised classification model. Which columns should be used as features?

   • ADistance, WeatherCondition, TrafficLevel
   • BDistance, WeatherCondition, Delayed
   • COrderID, Distance, WeatherCondition
   • DOrderID, TrafficLevel, Delayed
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A bank wants to predict loan default risk. They have a dataset with columns: CustomerID, AnnualIncome, CreditScore, ExistingDebt, and Defaulted (Yes/No). They plan to train a supervised classification model. Which set of columns should be used as features?

   • ACustomerID, AnnualIncome, CreditScore
   • BAnnualIncome, CreditScore, ExistingDebt
   • CAnnualIncome, CreditScore, Defaulted
   • DCustomerID, ExistingDebt, Defaulted
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A marketing agency wants to predict whether a user will click on an ad. They have a dataset with columns: UserID, Age, TimeSpentOnSite, NumberOfVisits, and Clicked (Yes/No). They plan to train a supervised classification model. Which columns should be used as features?

   • AUserID, NumberOfVisits, Clicked
   • BAge, TimeSpentOnSite, Clicked
   • CAge, TimeSpentOnSite, NumberOfVisits
   • DUserID, Age, TimeSpentOnSite
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A data analyst has a dataset of 1,000 customer records with a binary label indicating churn. They need to split the data so that the model can be trained and then evaluated on unseen data. Which approach best ensures a reliable evaluation?

   • ATrain on all 1,000 records and validate on a subset of the same records.
   • BUse the first 800 records for training and the last 200 for validation.
   • CUse 50% for training and 50% for validation, but only if the data is sorted by date.
   • DRandomly split 80% for training and 20% for validation.
   Reveal Solution Discussion Select an answer first