Class Imbalance

Question 1

What is class imbalance in machine learning

Answer 1

a situation where a dataset’s predictor variable contains more instances of one outcome than another.

Question 2

What are the majority and minority classes

Answer 2

refers to the class with more instances while the minority class has fewer instances.

Question 3

What are the two techniques to fix potential issues with class imbalance

Answer 3

upsampling and downsampling.

Question 4

What does downsampling do?

Answer 4

altering the majority class by using less of the original dataset to produce a more even split.

Question 5

What is upsampling?

Answer 5

It artificially increases the frequency of the minority class.

Question 6

How common is it for a dataset to have a perfectly balanced split of classes?

Answer 6

It’s extremely rare. Most datasets tend to have some degree of imbalance, where one class has significantly more examples than others.

Question 7

Can a dataset with some imbalance still be useful for training a machine learning model?

Answer 7

Absolutely! In many cases, a moderate imbalance like 70/30 or 80/20 is perfectly acceptable for training. The model can still learn effectively from the data

Question 8

When does class imbalance become a major concern for machine learning models?

Answer 8

Major issues arise when the majority class makes up 90% or more of the dataset. This can cause the model to become biased towards the majority class and perform poorly on the minority class

Question 9

For which variable should class imbalance be considered?

Answer 9

for categorical variables

Question 10

Which model is class balance applicable to?

Answer 10

classification models

Question 11

What is class balancing in machine learning?

Answer 11

Class balancing refers to techniques that adjust the number of samples in a dataset to make the proportions of different classes more even.

Question 12

Why is class balancing important?

Answer 12

Imbalanced datasets can lead models to be biased towards the majority class and perform poorly on the minority class.

Question 13

Which datasets is downsampling suitable for?

Answer 13

Downsampling is suitable for large datasets (tens of thousands of observations or more).

It’s important to ensure model performance doesn’t suffer due to reduced data.

Question 14

How is downsampling done?

Answer 14

By randomly selecting and removing observations from the majority class.

Question 15

When to use upsampling?

Answer 15

Upsampling is used for smaller datasets where removing data from the majority class is not feasible.

Question 16

Name 2 ways upsampling can be done?

Answer 16

There are two main methods:
1. Duplication: Simply copying existing minority class observations.
2. Synthetic Minority Oversampling Technique (SMOTE): Creates new, unique data points for the minority class based on existing data.

Question 17

Why is it important to keep a separate test set unaltered when applying downsampling or upsampling techniques?

Answer 17

1.Real-World Performance: We want to understand how well the model performs on unseen data that reflects the actual class distribution in the real world.
2. Avoiding Overfitting: Balancing techniques can lead to overfitting if applied to the test data. Overfitting occurs when the model memorizes the specific patterns in the training data and performs poorly on new, unseen data.

Question 18

What are consquences of class rebalancing?

Answer 18

1. Over-predicting minority class: Class balancing can lead a model to over-recognize the minority class during prediction. This happens because the model learns from a data distribution that’s different from the real world.
2. Impact on class probabilities: Techniques like downsampling and upsampling can affect the underlying probabilities the model learns for each class. This can be especially impactful for algorithms like Naive Bayes that rely on these probabilities.

Question 19

At what point should you consider doing a class rebalance?

Answer 19

Class rebalancing should be reserved for situations where other alternatives have been exhausted and you still are not achieving satisfactory model results.

Question 20

At what percentage is class imbalance severe enough to consider rebalancing?

Answer 20

• A moderate (< 20%) imbalance may not require any rebalancing.
• An extreme imbalance (< 1%) would be a more likely candidate.

Question 21

Why is it important to check class balance?

Answer 21

If there is a lot more representation of one class than another, then the model may be biased toward the majority class. When this happens, the predictions may be inaccurate.