Society of Actuaries (SOA) PA Practice Exam

What do modeling techniques like GLMs benefit from after applying undersampling or oversampling?

1. Increased complexity in the model structure

2. Improved accuracy in predicting minority class outcomes

3. Decreased computation time and resource allocation

4. Loss of information from majority class instances

The correct answer is: Improved accuracy in predicting minority class outcomes

Modeling techniques, such as Generalized Linear Models (GLMs), benefit significantly from data manipulation methods like undersampling and oversampling primarily through improved accuracy in predicting minority class outcomes. In many datasets, especially those involving binary classification, it's common to encounter a situation called class imbalance, where one class (often the minority) represents a significantly smaller proportion of the dataset than the other class (the majority). This imbalance can lead to biased models that perform well on the majority class but poorly on the minority class. By utilizing undersampling (reducing the number of instances of the majority class) or oversampling (increasing the number of instances of the minority class), the distribution of classes becomes more balanced. This adjustment helps machine learning algorithms, including GLMs, to learn more effectively about the minority class, as it ensures that the model has enough representative data to understand its characteristics. As a result, the model's ability to predict outcomes related to the minority class improves, leading to better overall accuracy and performance metrics for that class. In summary, oversampling and undersampling lead to a more equitable representation of classes in the training data, which is essential for enhancing the model's predictive accuracy, particularly concerning minority class outcomes.