How Data, Algorithms, and Models Form AI Foundations AIE Practice Questions (Page 3)

Questions 11–15

1. 11expert · hard

   A marketing team wants to build a model to predict customer lifetime value (CLV). They have a dataset with 1 million customers, but the CLV distribution is highly skewed: most customers have low CLV, a few have very high CLV. They train a linear regression model and find it performs poorly on high-CLV customers. What is the most effective approach?

   • ARemove the high-CLV customers from the dataset to make the distribution normal
   • BTransform the target variable (e.g., log transform) to reduce skewness before training
   • CTrain the model only on high-CLV customers
   • DUse a more complex linear model with more features
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A hospital wants to build a model to predict patient deterioration. They have electronic health records (EHR) with many missing values, but they also have real-time vital sign data. The team is debating whether to impute missing values or use an algorithm that handles missing data natively. What is the most important consideration?

   • AThe model should be trained only on complete cases to ensure accuracy
   • BThe algorithm should be chosen first, and data quality is secondary
   • CThe imputation method must preserve the clinical meaning of the data to avoid biased predictions
   • DReal-time data should be ignored because it is less reliable than EHR
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A logistics company wants to use AI to estimate delivery times based on distance, traffic, and weather. They have historical delivery records. Which approach best illustrates the role of an algorithm in this AI system?

   • ACreating a visual dashboard to display delivery metrics
   • BWriting a step-by-step procedure that the system follows to learn patterns from the historical records
   • CStoring the historical records in a database for future reference
   • DHiring data scientists to manually review each delivery record
   Reveal Solution Discussion Select an answer first
5. 14expert · hard

   A telecom company wants to predict customer churn. They have a dataset with 500,000 customers, but only 5% churned. They train a model and achieve 95% accuracy, but the model fails to identify most churners. What is the most likely cause?

   • AThe model is overfitting the training data
   • BThe algorithm is not powerful enough to learn churn patterns
   • CThe model is predicting the majority class (non-churn) most of the time, leading to high accuracy but poor recall for churn
   • DThe dataset is too large for the algorithm to process
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A city traffic department wants to predict congestion levels. They have sensor data from intersections and plan to use a regression algorithm. Which statement best describes how these components interact?

   • AThe algorithm learns from sensor data to create a model that predicts congestion
   • BThe model is used to collect sensor data for the algorithm
   • CThe algorithm and model are the same thing
   • DThe sensor data is the algorithm, and the model is the output
   Reveal Solution Discussion Select an answer first