BioHEL's Rule Post-processing Engine

The Rule Post-processing Engine is a standalone program which takes a set of rules generated with BioHEL and applies a series of memetic operators over the individual rules and the whole rule set to increase their generality and reduce their complexity.

Documentation

The operators used by the post-processing engine are described in detail in Franco2012a. Below we provide a brief summary:

• Rule cleaning - a rule operator that tries to reduce the space covered by the attributes to minimise the number of misclassified examples. There are two variations of this operator: conservative (deactivates attributes which appear only in negative examples) and non-conservative (deactivates all attributes which do not appear in positive examples).
• Rule pruning - a rule-wise operator that erases the unnecessary attributes in the rules. this is done by scanning the rule iteratively and attempting to remove the attributes that do not affect the accuracy of the rule.
• Rule swapping - a ruleset-wise operator that swaps the order of the rules according to their coverage to try to erase rules that are irrelevant. This operator helps reducing the cardinality of the final solutions.

Download

You can download C++ source code of the Rule Post-processing Engine.

Read the tutorial to learn how to install and use the engine.

Publications

If you use the Rule Post-processing Engine, please cite the following publication:

• DOI slides code BibTeX
  M.A. Franco, N. Krasnogor, J. Bacardit
  Post-processing operators for decision lists
  in Proceedings of the fourteenth international conference on Genetic and evolutionary computation conference - GECCO '12, p.847, Philadelphia, Pennsylvania, USA, 2012

  @INPROCEEDINGS{Franco2012a,
    title = {Post-processing operators for decision lists},
    author = {Franco, Maria A. and Krasnogor, Natalio and Bacardit, Jaume},
    year = 2012,
    doi = {10.1145/2330163.2330281},
    booktitle = {Proceedings of the fourteenth international conference on Genetic and evolutionary computation conference - GECCO '12},
    pages = {847},
    address = {Philadelphia, Pennsylvania, USA}
  }
  

We would also be happy to list your related publications on this website, so feel free to contact us at jaume.bacardit.