ML Workflow CT-AI Practice Questions (Page 3)

Questions 11–15

1. 11expert · hard

   A company has deployed a model to predict credit risk. The model was trained on data from the past five years. Recently, the economic environment has changed significantly, and the model's performance has degraded. The team has limited resources and must decide how to respond. Which approach is most appropriate?

   • AIncrease the model's complexity to capture the new patterns
   • BRetrain the model using recent data to adapt to the new economic conditions
   • CKeep the current model and manually adjust its outputs based on expert judgment
   • DReduce the model's sensitivity to changes by lowering its learning rate
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A data scientist is preparing a dataset for a model that predicts customer satisfaction. The dataset contains a feature 'age' with some extreme values (e.g., 200) that are likely data entry errors. What is the most appropriate action?

   • ATreat the extreme values as missing and impute them using a reasonable method
   • BRemove the entire 'age' feature from the dataset
   • CNormalize the entire feature to reduce the impact of extreme values
   • DKeep the extreme values as they are to preserve all data
   Reveal Solution Discussion Select an answer first
4. 13expert · hard

   A team is training a deep learning model for natural language processing. They have a limited amount of labeled data but a large amount of unlabeled text data. The model is overfitting. Which approach is most likely to improve generalization?

   • AIncrease the model's capacity to capture more patterns from the limited data
   • BTrain the model from scratch on the unlabeled data using unsupervised learning
   • CUse transfer learning from a pre-trained language model and fine-tune on the labeled data
   • DAdd more layers to the model to increase its depth
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A bank wants to build a model to approve loan applications. The data science team has collected historical loan data with customer demographics, credit scores, and loan outcomes. They plan to use a supervised classification algorithm. Which sequence of steps best represents the typical ML workflow they should follow?

   • AEvaluate model → collect data → clean data → train model → deploy
   • BCollect data → train model → clean data → evaluate model → deploy
   • CTrain model → collect data → evaluate model → deploy → clean data
   • DCollect data → clean and split data → train model → evaluate model → deploy and monitor
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A data scientist is building a model to predict housing prices. The dataset includes a categorical feature 'neighborhood' with 50 unique values. The scientist wants to use a linear regression model. What is the most appropriate way to handle this categorical feature?

   • AApply one-hot encoding to create binary columns for each neighborhood
   • BAssign integer codes to each neighborhood (1 to 50)
   • CUse label encoding and treat it as a continuous variable
   • DRemove the neighborhood feature entirely
   Reveal Solution Discussion Select an answer first