Dealing with imbalanced datasets is a very common classification problem.
Consider a Binary classification problem. Your goal is to predict a positive versus a negative class. The ratio between the two classes is highly skewed in favor of the positive class. This situation is frequently encountered in the following instance:
- In a medical context where the positive class corresponds to the presence of cancerous cells in people out of a large random population
- In a marketing context where the positive class corresponds to prospects buying an insurance while the majority of people are not buying it
In both these cases, we want to detect the samples in the minority class, but they are overwhelmingly outnumbered by the samples ...
