WhiBo is an open source platform for White Box (component based) machine learning algorithm design. It is intended to be used by typical end users, research scientists and algorithm developers. The main idea of WhiBo is to offer standardized components for algorithm design which will enable simple design and performance testing, easy extension of the component repository and creation of new generic algorithms.
Component based approach defines one generic algorithm for every class of problems (e.g. decision trees, partitioning clustering etc.). Every algorithm (problem) is divided into common sub-problems which could be solved with one ore more reusable components (RCs). RCs are extracted from existing algorithms and their partial improvements and are saved in a component repository. This structure enables reconstruction of algorithms that components originate from, as well as design of hybrid algorithms by the interchange of existing components, in one joint implementation.
WhiBo implements generic algorithms and RC repository and GUI for design of decision trees and partitioning clustering algorithms. Also an environment for testing algorithm performance and significance is provided.
WhiBo is implemented as Rapid Miner (version 5.x) plug-in. Installation instructions for Rapid Miner can be found on Rapid Miner's official website.
If you are using WhiBo or Component Based White-Box in scientific purposes please cite:
Delibašić B, Jovanović M, Vukićević M, Suknović M, Obradović Z (2011) Component-based decision trees for classification, Intelligent Data Analysis 15 (5), (accepted for publication), ISSN: 1088-467X
or
Delibašić B, Kirchner K, Ruhland J, Jovanović M, Vukićević M (2009) Reusable components for partitioning clustering algorithms, Artificial Intelligence Review, 32(1-4), p. 59-75 DOI: 10.1007/s10462-009-9133-6
