Question

Which of the following approaches can be used to correct for issues related to imbalanced classes?

A. Evaluating the model using Cohen’s kappa

B. Oversampling the minority class

C. Calibrating the model at the production stage

D. Collecting more data from the minority class

E. Using more robust performance metric

In recursive partitioning, how many homogeneous sub regions are created when two splits are introduced?

A. 3

B. 4

C. 2

In Recursive Partitioning scoring, we need to:

A. Create an aggregated scoring function based upon the mode of a regression model.

B. Define a function for each subregion

C. Create an aggregated scoring function based upon the mean of a classification model

D. Evaluate the estimated/predicted response for each subregion using an aggregated function

Which of the following statements are accurate in relation to univariate splitting for continuous variables?

A. Splits can depend on any transformation applied to the variables

B. It accounts well for irregularities in data

C. It defines subregions of space in rectangular boxes

D. Splits can happen in a perpendicular fashion along x and y axes.

When splitting categorical variables along important sets of features, it is best to:

A. Start building a tree in independent categories method first and then proceed to a group categories method

B. Use group categories method only to split along predefined features

C. Start with group categories method and proceed to independent categories to increase precision

D. Build a model that favors categorical variables with a lot of categories

Answer 1

ANSWER:

Imbalanced data typically refers to a problem with classification problems where the classes are not represented equally.

• Oversampling the minority class : Oversampling can be defined as adding more copies of the minority class. Oversampling can be a good choice when you don’t have a ton of data to work with.
• Evaluating the model using Cohen's Kappa : Classification accuracy normalized by the imbalance of the classes in the data.
• Using more robust performance metric : Accuracy is not the metric to use when working with an imbalanced dataset.

1. Confusion Matrix
2. Precision
3. Recall
4. F1 Score.

I tried my best to answer. I hope you understood.