Question 1

Using the following chart for age and income, determine the split points for income.

Accepted Answer

The answer of Using the following chart for age and...

Question 2

Which is not a purpose of running classification and regression trees (CART)?

Accepted Answer

CART is primarily used for decision making by splitting data into subsets, which improves prediction accuracy and helps in understanding the data features that contribute to the outcome. It does not aim to identify the most diverse case set for the target variable, but rather to create a model that can accurately classify or predict outcomes for new data. Options A, B, and D are all purposes of running CART, as they relate to improving the model's performance and interpretability.

Question 3

If the RMSE for the validation set is 56.91 and the RMSE for the test set is 55.39, then what range will the new data RMSE lie in?

Accepted Answer

The answer of If the RMSE for the validation set...

Question 4

If the RMSE for the validation set is 58.78 and the RMSE for the test set is 57.12, then what range will the new data RMSE lie in?

Accepted Answer

The answer of If the RMSE for the validation set...

Question 5

A regression tree was developed to predict customer spending for a hotel during football season. One of the leaf nodes consists of six cases in the training set with the following values: 312.00, 350.00, 285.00, 295.00, 423.00, 249.00. What is the predicted spending amount on a hotel for the night for a customer that falls into this leaf node?

Accepted Answer

The answer of A regression tree was developed to predict...

Question 6

A regression tree was developed to predict customer spending for a hotel during football season. One of the leaf nodes consists of six cases in the training set with the following values: 312.00, 350.00, 285.00, 295.00, 380.00, 220.00. What is the predicted spending amount on a hotel for the night for a customer that falls into this leaf node?

Accepted Answer

The predicted spending amount is the mean of the values in the leaf node: (312.00 + 350.00 + 285.00 + 295.00 + 380.00 + 220.00) / 6 = 307.00.

Question 7

When using the CART algorithm, the Gini index is used in the classification tree, however in a regression tree, _____ is used to measure impurity.

Accepted Answer

In a regression tree, the mean squared error (MSE) is used to measure impurity. This is because MSE provides a clear and differentiable metric for evaluating the variance of the continuous target variables within each node, which is essential for optimizing the tree structure.

Question 8

Using the following sample of a regression prune log, the minimum error tree is decision node # 19 with a standard error of 4.689492 (not shown). Using the information provided, which decision node number represents the best-pruned tree?

Accepted Answer

The answer of Using the following sample of a regression...

Question 9

The following table reflects a partial Analytic Solver's Performance measure for a hotel cost during an NFL game night. What is the MAD implying?

Accepted Answer

The answer of The following table reflects a partial Analytic...

Question 10

The following table reflects a partial Analytic Solver's Performance measure for a hotel cost during an NFL game night. What is the MAD implying?

Accepted Answer

The answer of The following table reflects a partial Analytic...

Question 11

When generating a single regression tree visually, the prp function is used. Based on the following example code, what does setting type = 1 mean?
>prp(default_tree, type = 1, extra = 1, under = TRUE)

Accepted Answer

The answer of When generating a single regression tree visually,...

Question 12

Using the following pruning table, which tree is the best-pruned tree?

Accepted Answer

The answer of Using the following pruning table, which tree...

Question 13

Which option is not one of the three common strategies used in creating ensemble models?

Accepted Answer

Bootstrapping is a statistical method used to sample data, but it is not considered a standalone strategy for creating ensemble models. Bagging, boosting, and Random Forest are specific ensemble techniques.

Question 14

If the performance measures are based on a cutoff value of 0.5, then if we lower the cutoff value, more cases will be in the target class, resulting in different performance measurement values. What chart can be used to review the data that are independent of the cutoff value?

Accepted Answer

All options are independent of the cutoff value.

Question 15

If predictor variables are highly correlated, then repeated sampling of the training data and a random selection of features are used to construct trees. This is an example of which strategy?

Accepted Answer

Random Forest utilizes both repeated sampling of the training data (bootstrap aggregating or bagging) and random selection of features to construct multiple decision trees, aiming to reduce overfitting and improve model accuracy.

Question 16

In a random forest model, as a guideline the user needs to select a number of the random features for each tree. If there are 196 predictor variables in the data, each tree will randomly select how many features to be included in the tree?

Accepted Answer

The rule of thumb for selecting the number of features in each split in a random forest model for classification problems is the square root of the total number of features. In this case, the square root of 196 is 14.

Question 17

In a random forest model, as a guideline the user needs to select a number of the random features for each tree. If there are 9 predictor variables in the data, each tree will randomly select how many features to be included in the tree?

Accepted Answer

In a random forest model, a good rule of thumb for classification problems is to use the square root of the total number of features. Since there are 9 predictor variables, the square root of 9 is 3. Therefore, each tree would randomly select 3 features to be included.

Question 18

When constructing the argument for a bagging tree strategy, the varImpPlot function displays feature importance graphically. For this we set the type argument to either equal 1 or 2. If type = 2, then what does this command?

Accepted Answer

The `varImpPlot` function with `type = 2` uses the average decrease in the Gini impurity index to compare the feature importance. This is a common method for measuring how each feature contributes to the homogeneity of the nodes and leaves in the decision trees within the ensemble.

Question 19

Ensemble tree models combine multiple single-tree models to reduce the variation in prediction error. Of the strategies, which may lead to overfitting?

Accepted Answer

Boosting can lead to overfitting because it focuses on correcting the mistakes of previous models, potentially leading to a model that is too complex and closely fitted to the training data.

Question 20

In the following tree, how many leaf nodes are there?

Accepted Answer

The answer of In the following tree, how many leaf...

Quiz 10: Supervised Data Mining: Decision Trees

Using the following chart for age and income, determine the split points for income.

Which is not a purpose of running classification and regression trees (CART) ?

If the RMSE for the validation set is 56.91 and the RMSE for the test set is 55.39, then what range will the new data RMSE lie in?

If the RMSE for the validation set is 58.78 and the RMSE for the test set is 57.12, then what range will the new data RMSE lie in?

When using the CART algorithm, the Gini index is used in the classification tree, however in a regression tree, _____ is used to measure impurity.

Using the following sample of a regression prune log, the minimum error tree is decision node # 19 with a standard error of 4.689492 (not shown) . Using the information provided, which decision node number represents the best-pruned tree?

The following table reflects a partial Analytic Solver's Performance measure for a hotel cost during an NFL game night. What is the MAD implying?

The following table reflects a partial Analytic Solver's Performance measure for a hotel cost during an NFL game night. What is the MAD implying?

When generating a single regression tree visually, the prp function is used. Based on the following example code, what does setting type = 1 mean? >prp(default_tree, type = 1, extra = 1, under = TRUE)

Using the following pruning table, which tree is the best-pruned tree?

Which option is not one of the three common strategies used in creating ensemble models?

If the performance measures are based on a cutoff value of 0.5, then if we lower the cutoff value, more cases will be in the target class, resulting in different performance measurement values. What chart can be used to review the data that are independent of the cutoff value?

If predictor variables are highly correlated, then repeated sampling of the training data and a random selection of features are used to construct trees. This is an example of which strategy?

In a random forest model, as a guideline the user needs to select a number of the random features for each tree. If there are 196 predictor variables in the data, each tree will randomly select how many features to be included in the tree?

In a random forest model, as a guideline the user needs to select a number of the random features for each tree. If there are 9 predictor variables in the data, each tree will randomly select how many features to be included in the tree?

When constructing the argument for a bagging tree strategy, the varImpPlot function displays feature importance graphically. For this we set the type argument to either equal 1 or 2. If type = 2, then what does this command?

Ensemble tree models combine multiple single-tree models to reduce the variation in prediction error. Of the strategies, which may lead to overfitting?

In the following tree, how many leaf nodes are there?