Why was model-building strategy #1 used to estimate generalization performance on the income data for the decision tree, logistic regression, and random forest models?

Question

BlackTom AI · Accepted Answer

Start by identifying what the question is asking: why was model-building strategy #1 chosen to estimate generalization performance for the income data across three models (decision tree, logistic regression, random forest).
Option 1: 'There were 48,842 rows of data, so nested cross validation would have taken a long time to run' — This reflects a practical concern about computational cost. Nested cross-validation is more intensive because it involves an inner loop for hyperparameter tuning and an outer loop for performance estimation. When the dataset is large (e.g., tens of thousands of rows), nested CV can be prohibitively slow, which is a common reason to opt for a simpler strategy. The statement correctly points to computational feasibility as a reason to avoid nested CV in this context.
Option 2: 'Model-building strategy #1 leads to a better estimate of generalization performance, because the estimate is based on more splits of the data' — More splits can sometimes stabilize estimates, but more splits alone do not guarantee a better estimate. Nested CV, by contrast, provides an almost unbiased estimate by combining hyperparameter tuning with outer validation in a principled way. If strategy #1 is not nested CV, claiming it inherently gives a better estimate due to more splits is not generally correct and depends on how the splits are used. This makes the assertion dubious without additional context.
Option 3: 'Model-building strategy #1 is more appropriate than model-building strategy #2 for logistic regression and decision tree models' — This kind of claim would depend on how each strategy handles model types. It is not automatically true that one strategy is universally more appropriate for both logistic regression and decision trees; different models interact differently with data splitting, hyperparameter tuning, and class imbalance. Without specifics, this blanket claim is not a robust justification.
Option 4: 'In Model-building strategy #1 the splits are always stratified, which is the best approach for classification models' — Stratified splitting preserves class proportions, which is important for classification with imbalanced classes, but the statement asserting it is the 'best approach' across all cases is too strong. Some scenarios may not require stratification, and 'always' stratified splits may not be necessary or optimal depending on data and validation goals. Therefore, as a general justification, this is an overreach unless strategy #1 explicitly guarantees stratification and there is a demonstrated need for it in this dataset.
In summary, the most defensible reasoning among the options hinges on practical computational considerations given the data size, which aligns with option 1. The other options make claims about general superiority, model-type suitability, or universal use of stratification that lack sufficient support without more context or explicit details about Model-building strategy #1.

Why was model-building strategy #1 used to estimate generalization performance on the income data for the decision tree, logistic regression, and random forest models?单项选择题

类似问题

What is the purpose of cross validation?

Among the following training and evaluation strategies, which of them divides the dataset into k subsets and, at each iteration, a different subset is reserved for testing and all the others, used for training?

What is the purpose of performing cross-validation?

When training models using machine learning, a very powerful technique for tuning and optimizing the best model is:

In k-fold cross-validation, what happens in each iteration?

What is one advantage of k-fold cross-validation over a single holdout set?

Which of the following is TRUE about cross validation?

更多留学生实用工具

智能学习助手

风格化写作助手

论文查重助手

文献引用助手

课堂转译助手

课堂笔记助手

Quiz搜索助手

学校历年真题

智能刷题助手

智能匹配练习

希望你的学习变得更简单