Pusion (Python Universal Fusion) is a generic and flexible framework written in Python for combining multiple classifier’s decision outcomes.

The framework comprises a variety of decision fusion methods adapted from the field of ensemble techniques and pattern recognition. The general purpose of the framework is to improve classification performance over input classifiers and to determine the compatibility of individual decision fusion methods for a given problem. Pusion handles an unlimited number of classifiers and also tolerates various forms of classification output. These include multiclass and multilabel classification, as well as crisp and continuous class assignments (see documentation for more details). The framework was originally designed for combining fault detection and fault diagnosis methods, but its application is not limited to this domain.

The full HTML documentation is available on GitHub Pages or in the project's docs/build/ folder.