Model Selection and Evaluation MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 2)

Questions 6–10

1. 6expert · hard

   A data science team is developing a model to predict equipment failure in a manufacturing plant. They have a dataset with 5,000 samples and 200 features. The team is considering three models: a linear regression, a decision tree with unlimited depth, and a random forest with 100 trees. The linear regression achieves a training RMSE of 50 and a test RMSE of 55. The decision tree achieves a training RMSE of 10 and a test RMSE of 80. The random forest achieves a training RMSE of 20 and a test RMSE of 45. Which model should they choose?

   • ALinear regression, because it is the simplest model and less likely to overfit.
   • BRandom forest, because it has the lowest test RMSE.
   • CDecision tree, because it has the lowest training RMSE.
   • DLinear regression, because it has the smallest gap between training and test RMSE.
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   A credit card fraud detection model is being developed. The cost of a false negative (missing a fraudulent transaction) is 100 times higher than the cost of a false positive (flagging a legitimate transaction). Which metric should be prioritized during model selection?

   • AAUC
   • BRecall
   • CAccuracy
   • DPrecision
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A data scientist is predicting house prices. They log-transform the target variable (price) to reduce skewness and train a linear regression model. The model's RMSE on the log-transformed target is 0.15. What is the RMSE in the original price scale?

   • A0.15
   • Bexp(0.15) ≈ 1.16
   • CCannot be determined without the original data
   • Dexp(0.15) - 1 ≈ 0.16
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A credit card company is building a model to approve or decline transactions. The cost of a false decline (blocking a legitimate purchase) is customer dissatisfaction, while the cost of a false approval (allowing fraud) is financial loss. The company wants to minimize the total cost, but the costs are not equal. Which metric should they use to evaluate the model?

   • AROC-AUC
   • BF1-score
   • CAccuracy
   • DCost-sensitive metric (e.g., weighted error rate)
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A data scientist is building a pipeline to predict house prices. They have a dataset with many categorical features (e.g., neighborhood, house style) and numeric features (e.g., square footage, lot size). They need to transform the categorical features into a format suitable for a linear regression model. Which approach correctly uses a transformer in the pipeline?

   • AUse LinearRegression to transform the categorical features, then fit a OneHotEncoder on the output.
   • BUse OneHotEncoder to fit the model directly on the raw categorical features.
   • CUse LabelEncoder to convert categorical features to ordinal integers, then fit a LinearRegression on the transformed data.
   • DUse OneHotEncoder to convert categorical features to binary columns, then fit a LinearRegression on the transformed data.
   Reveal Solution Discussion Select an answer first