WhiBo brings better transparency, reproducibility, adaptation

The RapidMiner plugin for White-box algorithm design and experimentation.

It is friendly, extendable and free.
Try it for yourself.
Currently available for Decision trees (soon also for Representative based clustering)

Choose algorithm parts

Each algorithm can be thought of as a composition of subproblems that needs to be solved. The WhiBo framework allows users to pick among many components that solve each subproblem. This enables users to customize the algoritms for the task at hand. Additionally, each component can be fine-tuned by its parameters. And all this through a friendly graphical user interface, within RapidMiner.

Experiment with Decision Tree and Clustering algorithms

Currently, WhiBo offers components for assembling decision tree, and (soon!) representative-based clustering algorithms. Use WhiBo to analyze what components are responsible for good/bad performance. Learn the parts of algorithms without going into too many details.

Automatic search for good components

Relax while an evolutionary algorithm searches through the algorithm-space, in order to find the best performing composition for the given data. Set the evolutionary strategy parameters, follow the performance evolution in detailed logs, and learn which components bring the survival of the fittest algorithm.

Installation and usage

WhiBo plugin is currently developed for RapidMiner 6.
Stay tuned for updates by pressing Join in top right corner.

You can install WhiBo plugin using RapidMiner Marketplace, which is located in Menu bar under Help menu. After opening RapidMiner Marketplace please enter in search box WhiBo and press Search. Once located, please check Select for install/update and press Install x packages. Please read and accept the terms of license and press Install x packages.

If you are interested in using WhiBo package without any background please consult User Guide. However, we highly recommend User Manual where you can find complete tutorial with theoretical description of white box algorithms. If you need help setting up WhiBo RapidMiner process feel free to look at our demo processes on this link and this link.

Finnaly, if you are interested in developing and further extension of WhiBo package you can consult Developer Guide and you can see our GitHub page.

Research findings

From 2008, numerous papers have shown benefits of white box algorithm design.

[We] showed that using RC based algorithms can allow data mining practitioners to achieve better adaptation for data at hand... One of the advantages of the component-based design approach is that it gives further possibilities for more detailed analysis of RC individual and synergic performance on a dataset which is analyzed in the following subsections.

Delibasic et al. 2012

Data & Knowledge Eng.
...experiments showed that reuse and interchange of reusable components (RCs) extracted from the existing algorithms can lead to better results compared to the algorithms they originate from... Further, single algorithms composed from RCs showed better performance than consensus clustering frameworks that are using traditional single algorithms (e.g., K-means, SOM)

Vukicevic et al, 2013

Knowledge and Inf. Sys.
...large algorithmic space is hard to search since the objective functions are computationally very expensive... Experimental evidence show that utilization of evolutionary algorithms for design of well performing component-based algorithms is useful because it automates the process of ﬁnding the appropriate algorithm for data at hand by ﬁnding the right component interplay and by adjusting parameters of the assembled components.

Jovanovic et al, 2014

Intelligent Data Analysis
[enables] easier development of new decision tree based algorithms that are constructed using RCs found in existing and well known algorithms... generic framework for component-based algorithms design enhances understanding, testing and usability of decision tree algorithm parts.

Suknovic et al, 2012

Computational Statistics

Full list of references:

Vukicevic, M., Radovanovic, S., Delibasic, B., Suknovic, M. (2016). Extending meta-learning framework for clustering gene expression data with component-based algorithm design and internal evaluation measures. International Journal of Data Mining and Bioinformatics, 14(2), 101-119. 2016
Vukicevic M., Radovanovic S., Delibasic B., Suknovic M. (2016) White-Box Predictive Algorithms for Predicting Disease, States on Gene Expression Data – From Component Based Design to Meta Learning. In Proc. of 2016 Belgrade Bioinformatics Conference, Belgrade, Serbia.2016
Jovanovic M, Delibasic B, Vukicevic M, Suknovic M, Martic M (2014), Evolutionary approach for automated component-based decision tree algorithm design, Intelligent Data Analysis 18(1), 25-42.2014
Vukićević, M., Radovanović, S., Milovanović, M., Minović, M. (2014). Cloud Based Metalearning System for Predictive Modeling of Biomedical Data. The Scientific World Journal, 2014, Article ID 859279, 10 pages.2014
Vukicevic M, Kirchner K, Delibasic B, Jovanovic M, Ruhland J, Suknovic M (2013) Finding best algorithmic components for clustering microarray data, Knowledge and Information Systems,https://doi.org/10.1007/s10115-012-0542-5.2013
B. Delibasic, M. Vukicevic, M. Jovanovic, M. Suknovic (2013) White-box decision tree algorithms: A pilot study on perceived usefulness, perceived ease of use, and perceived understanding, International Journal of Engineering Education 29 (3), p. 674–687.2013
Suknovic M, Delibasic B, Jovanovic M, Vukicevic M, Becejski-Vujaklija D, Obradovic Z (2012) Reusable Components in Decision Tree Induction Algorithms, Computational Statistics. DOI: https://doi.org/10.1007/s00180-011-0242-8.2012
B. Delibasic, M. Vukicevic, M. Jovanovic, K. Kirchner, J. Ruhland, M. Suknovic (2012) An architecture for component-based design of representative-based clustering algorithms, Data & Knowledge Engineering. doi: https://doi.org/10.1016/j.datak.2012.03.005.2012
M. Vukicevic, M. Jovanovic, B. Delibasic, S. Isljamovic, M. Suknovic (2012) Reusable component-based architecture for decision tree algorithm design, International Journal on Artificial Intelligence Tools. doi: https://doi.org/10.1142/S0218213012500224.2012
Vukicevic M, Delibasic B, Obradovic Z, Jovanovic M, Suknovic M (2012), A Method for Design of Data-tailored Partitioning Algorithms for Optimizing the Number of Clusters in Microarray Analysis, 2012 IEEE Symposium on Computational Intelligence in Bioinformatics and Computational Biology, doi: https://doi.org/10.1109/CIBCB.2012.6217238.2012
Delibasic B, Jovanovic M, Vukicevic M, Suknovic M, Obradovic Z (2011) Component-based decision trees for classification, Intelligent Data Analysis 15(5), 671-693.2011
Jovanović Miloš, Delibašić Boris, Vukićević Milan, Suknović Milija (2011) Optimizing performance of decision tree component-based algorithms using evolutionary algorithm in RapidMiner, In proc. of the 2nd RapidMiner Community Meeting and Conference, June 7-10, Dublin, Ireland, www.rcomm2011.org, 135-149.2011
Delibašić B, Jovanović M, Vukićević M, Suknović M, Kathrin Kirchner, Johannes Ruhland, Zoran Obradović (2011) A decision support system architecture for data mining based on reusable components (patterns), In digital proceedings of the EWG-DSS London 2011 Workshop on Decision Support Systems, June 23-24, London, UK, 35.2011
Vukićević Milan, Jovanović Miloš, Delibašić Boris, Suknović Milija (2010) WhiBo - RapidMiner plug-in for component based data mining algorithm design, In proc. of the 1st RapidMiner Community Meeting and Conference, September 13-16, Dortmund, Germany, www.rcomm2010.org, p. 30-35.2010
Delibašić B, Kirchner K, Ruhland J (2010) Component-based software for clustering in data mining - A prototype in Matlab, SYMORG, Fakultet organizacionih nauka, Zlatibor, Zbornik radova izdat u CD formatu, ISBN: 978-86-7680-216-6.2010
Delibasic B, Kirchner K, Ruhland J, Jovanovic M, Vukicevic M (2009) Reusable components for partitioning clustering algorithms, Artificial Intelligence Review 32 (1), 59-75.2009