On Rule Learning Methods: A Comparative Analysis of Classic and Fuzzy Approaches
Classification is an important task widely researched by the machine learning and fuzzy communities. In this paper, we present and compare methods from both communities, in order to support the selection of a suitable method, according to two conflicting objectives: accuracy × interpretability. Two groups of rule-based methods are analysed: decision tree-based and genetic-based approaches. For the tree-based approaches, C4.5, PART and FuzzyDT, a fuzzy version of the C4.5 algorithm, are used. For the genetic-based approaches, MPLCS, a method from the machine learning community to generate rule-based models, SLAVE and FCA-Based, both fuzzy-based, are analysed. Since accuracy and interpretability are usually conflicting objectives, in this paper, we briefly present these methods and then discuss the models generated by them. Comparisons take into account the error rates and syntactic complexity of the produced models. Ten benchmark datasets are used in the experiments with a 10 fold cross-validation strategy. Results show that FCA-Based and MPLCS are able to obtain good accuracy and interpretability.
KeywordsFuzzy Systems Decision Trees Genetic Algorithms
Unable to display preview. Download preview PDF.
- 1.Quinlan, J.R.: C4.5: Programs for Machine Learning (Morgan Kaufmann Series in Machine Learning), 1st edn. Morgan Kaufmann (January 1993)Google Scholar
- 2.Quinlan, J.R.: Bagging, Boosting and C4.5. In: Proc. of the 13th Conf. on Artificial Intelligence, pp. 725–730 (1996)Google Scholar
- 3.Frank, E., Witten, I.H.: Generating accurate rule sets without global optimization. In: ICML 1998: Proceedings of the 15th Int. Conf. on Machine Learning, pp. 144–151. Morgan Kaufmann (1998)Google Scholar
- 7.Cintra, M.E., Camargo, H.A.: Fuzzy rules generation using genetic algorithms with self-adaptive selection. In: IEEE International Conference on Information Reuse and Integration - IRI, pp. 261–266 (2007)Google Scholar
- 8.Cintra, M.E., Monard, M.C., Camargo, H.A., Martin, T.P.: A hybrid approach for the automatic generation of fuzzy systems using fuzzy formal concepts. In: 2012 IEEE International Conference on Fuzzy Systems, FUZZ-IEEE 2012 (accepted for publication, 2012)Google Scholar
- 10.Cintra, M.E., Monard, M.C., Camargo, H.A.: An evaluation of rule-based classification models induced by a fuzzy method and two classic learning algorithms. In: The Brazilian Symposium on Artificial Neural Network (SBRN), vol. 1, pp. 188–193 (2010)Google Scholar
- 18.Mitchell, T.M.: Machine Learning. McGraw-Hill (1997)Google Scholar
- 19.Wille, R.: Restructuring lattice theory: An approach based on hierarchies of concepts. In: Rivals, I. (ed.) Ordered Sets, vol. 23, pp. 445–470 (1982)Google Scholar
- 23.Frank, A., Asuncion, A.: UCI machine learning repository (2010)Google Scholar
- 24.Bull, L., Bernadó-Mansilla, E., Holmes, J.: Learning Classifier Systems in Data Mining. Springer (2008)Google Scholar