multi-imbalance is an open-source Python library for multi-class imbalanced classification, which equips constantly growing Python community of machine learning researchers and practitioners with proper tools to deal with multi-class imbalance problems. The package follows the code conventions of popular sklearn package and provides efficient implementations of state-of-the-art decomposition techniques as well as both novel and classic resampling approaches for multi-class imbalanced classification. The code of the library is rigorously covered by unit tests, documented with sphinx library, and meets PEP8 standard. The code is available on GitHub
The package contains the implementation of the following methods:
Learn more from the video:
Our package can be easily installed with just one command:
pip install multi-imbalance
Below we present an exemplary usage of SOUP to resample the training data. We present the result of this preprocessing. The dataset we have used is Glass.
Below we present usage of SOUP along with OVO ensemble. Standard decision tree classifier from sklearn was used as component binary classifier.
As one can see, the usage of our library is very intuitive.