Foundational Concepts DY0-001 Practice Questions (Page 5)

Questions 21–24

1. 21application · medium

   A team has trained several different models (logistic regression, decision tree, and SVM) for a binary classification task. Each model has similar accuracy around 85%, but they make different errors on different samples. The team wants to improve the overall accuracy and robustness of the predictions. Which ensemble technique is most appropriate?

   • AVoting ensemble that combines the predictions of all three models
   • BBoosting with a single model type, such as decision trees
   • CSelecting the single best model based on validation accuracy
   • DBagging with a single model type, such as decision trees
   Reveal Solution Discussion Select an answer first
2. 22expert · hard

   A data scientist is building a model to predict customer lifetime value (CLV). The dataset includes historical transaction data. The data scientist uses a time-based split for training and validation: the first 80% of the data (by date) is used for training, and the last 20% is used for validation. The model performs well on the validation set. However, when deployed, the model's performance degrades over time. Which is the most likely cause?

   • AThe validation set was not large enough to provide a reliable estimate of performance.
   • BThe model is suffering from data leakage because the validation set is from a later time period.
   • CThe model is not being retrained, and the underlying data distribution is changing over time (concept drift).
   • DThe model is overfitting to the training data and does not generalize to new data.
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   What is the purpose of hyperparameter tuning in machine learning?

   • ATo find the optimal set of hyperparameters that improves model performance
   • BTo reduce the number of features in the dataset
   • CTo increase the size of the training dataset
   • DTo automatically clean the training data
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   A data scientist is building a model to predict loan default. The dataset contains 100,000 rows and 20 features. The data scientist uses a pipeline that includes a feature selection step (e.g., SelectKBest) and a classifier. They use a 10-fold cross-validation to evaluate the model. Which practice is most likely to cause data leakage?

   • APerforming feature selection on the entire dataset before splitting into folds
   • BScaling the features using a standard scaler fitted on the training folds only
   • CUsing a stratified k-fold cross-validation to maintain the class distribution in each fold
   • DUsing a pipeline in scikit-learn that integrates the feature selection and the classifier
   Reveal Solution Discussion Select an answer first