Rumble
Contact: Stefan Kramer
Categories: Prediction
Exposed methods:
predict | |
---|---|
Input: | Instances, feature vectors, class values |
Output: | Classification model |
Input format: | Weka's ARFF format plus text; Soon XML |
Output format: | Plain text |
User-specified parameters: | Norm used for learning Bound constant |
Reporting information: | Performance measures (sensitivity, specificity, AUC, prediction accuracy) |
Description:
RUMBLE (RUle and Margin Based LEarner) is a statistically motivated rule learning system based on the Margin
Minus Variance (MMV) optimization criterion [RUE08]. It can be adapted flexibly to a given dataset: First,
different types of data (structures, physico-chemical properties, logical background knowledge, ...) can be
handled by different plug-ins of the system (e.g., FTM plugin, Prolog plugin, Meta plugin, ...). Second, the
learning algorithm can be adapted to the noise level in the data by two regularization parameters. The main
algorithm performs a forward selection of variables as for linear or logistic regression models. The models
learned by RUMBLE are linear classifiers, i.e., they provide a linear weighting of the input features.
The software is implemented in the C++ programming language and was developed for the Linux and Mac OS
X operating systems. The RUMBLE software is dependent on the OpenBabel (http://www.openbabel.org)
chemistry toolbox. In case the Prolog plugin is used, there is also a dependency on the specific Prolog system
used. RUMBLE provides no graphical user interface (GUI) and is executed via the command line. The input
format accepted at the moment is Weka's [WIT99] ARFF format. XML input is under development. RUMBLE's
output is plain text.
For further information, we refer to the original publication [RUE08] and the website
http://wwwkramer.in.tum.de/research/machine_learning/margin_based
Background (publication date, popularity/level of familiarity, rationale of approach, further comments)
Published 2006-2008, best theory paper award at ILP 2006. Adopts the concept of a
margin from the Support Vector Machine (SVM), but focuses on the selection of
features instead of the selection of instances. Does not use kernels. Useful tool with
regularization parameter for noise handling and plug-ins for various data types (e.g.,
chemical structures and quantitative descriptors)
Bias (instance-selection bias, feature-selection bias, combined instance-selection/feature-selection bias, independence assumptions?, ...)
Feature-selection bias
Lazy learning/eager learning
Eager learning
Interpretability of models (black box model?, ...)
Good (linear classifier, i.e., produces a simple linear weighting of given features)
Type of Descriptor:
Interfaces:
Priority: Low
Development status: Development
Homepage: http://wwwkramer.in.tum.de/research/machine_learning/margin_based
Dependencies:
External components: OpenBabel
Technical details
Data: No
Software: Yes
Programming language(s): C++
Operating system(s): Linux
Input format: Instances, feature vectors, class values
Output format: Plain text
License: GPL
References
References:
[RUE08] Rückert, U and Kramer, S (2008). Margin-Based First-Order Rule Learning, Machine Learning, 70(2-3):189-206.
[WIT99] Witten, I.H. Frank, E., Data Mining: Practical Machine Learning Tools and Techniques with Java Implementations (Morgan Kaufmann, 1999).