Definition
BRF (Balance random forest) is a variant of random forest technique designed to solve the problem of imbalanced dataset. That is, a dataset in which a class has much more example of another one (e.g., in a dataset of bank transactions, regular transactions are much more frequent that fraudulent transactions).
It is a machine learning technique used to classify an input into a predetermined output class.
When to use
- when is important to include even the less represented class in a dataset (rare classes detection), reducing the bias toward more numerous classes
- in all the situations in which you need to detect rare cases (fraud detection, medical diagnostics, errors or anomalies in a SUT (system under test))
References
- Used in: (Kim, Wang, Kwon, et al., 2023)