Revisiting UCS: Description, Fitness Sharing, and Comparison with XCS

Orriols-Puig, Albert; Bernadó-Mansilla, Ester

doi:10.1007/978-3-540-88138-4_6

Cited by 32 publications

(48 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The first one, primarily aimed at studying performance through predictive accuracy, involves training SS-LCS and UCS with and without clustering-based initialization (SS-LCS CI / SS-LCS NI and UCS CI /UCS NI , respectively), as well as their rival algorithms in a variety of real-world classification problems. The performance metric used throughout this set of experiments for algorithm comparisons is the average accuracy rate of 5 tenfold stratified cross validation runs, in line with other comparative studies in the literature (Orriols-Puig et al 2008b;García et al 2009). …”

Section: Experimental Methodologymentioning

confidence: 97%

“…Methods that belong to the second approach and may provide an effective and computationally feasible alternative, in line with our initial requirement for high interpretability, include (a) classifiers inducing sets of rules, such as FOIL (Quinlan 1996), PART (Frank and Witten 1998), HIDER (Aguilar-Ruiz et al 2003, and SIA (Venturini 1993); (b) learning classifier systems, such as XCS (Wilson 1995), UCS (Bernadó-Mansilla and GarrellGuiu 2003;Orriols-Puig and Bernadó-Mansilla 2008b), GAssist (Bacardit 2004) and ILGA (Guan and Zhu 2005); and (c) algorithms inducing decision trees, such as C4.5 (Quinlan 1993).…”

Section: Introductionmentioning

confidence: 97%

“…The GA works on classifier conditions in an effort to adequately decompose the target problem into a set of subproblems, while supervised learning evaluates classifiers in each of them (Lanzi 2008). Although the most prominent example of this class of systems is UCS (Bernadó-Mansilla and Garrell-Guiu 2003;Orriols-Puig and Bernadó-Mansilla 2008b), an accuracybased LCS, we have recently introduced SS-LCS, a supervised strength-based LCS that departs from the accuracy-based approach to fitness computation (Tzima and Mitkas 2010). SS-LCS provides an efficient and robust alternative for offline classification tasks by extending previous strength-based frameworks (Wilson 1994;Bonelli et al 1990;Kovacs 2002a;b) and, together with UCS, will serve as the basis for our current investigation.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Clustering-based initialization of Learning Classifier Systems

2012

View full text Add to dashboard Cite

The present paper investigates whether an ''informed'' initialization process can help supervised LCS algorithms evolve rulesets with better characteristics, including greater predictive accuracy, shorter training times, and/or more compact knowledge representations. Inspired by previous research suggesting that the initialization phase of evolutionary algorithms may have a considerable impact on their convergence speed and the quality of the achieved solutions, we present an initialization method for the class of supervised Learning Classifier Systems (LCS) that extracts information about the structure of studied problems through a pre-training clustering phase and exploits this information by transforming it into rules suitable for the initialization of the learning process. The effectiveness of our approach is evaluated through an extensive experimental phase, involving a variety of realworld classification tasks. Obtained results suggest that clustering-based initialization can indeed improve the predictive accuracy, as well as the interpretability of the induced knowledge representations, and paves the way for further investigations of the potential of better-than-random initialization methods for LCS algorithms.

show abstract

Section: Experimental Methodologymentioning

confidence: 97%

Section: Introductionmentioning

confidence: 97%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Clustering-based initialization of Learning Classifier Systems

2012

View full text Add to dashboard Cite

show abstract

“…The reason is that it evolves only those highly-rewarded classifiers of the match set in the correct set, which predict the same class as that of the training example [33]. In comparison, GAssist has serious problems in dealing with multi-class problems -specially when the number of output classes are more than 5.…”

Section: Effect Of Nature Of Datasetmentioning

confidence: 99%

Classification Potential vs. Classification Accuracy: A Comprehensive Study of Evolutionary Algorithms with Biomedical Datasets

Tanwani

Farooq

2010

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Biomedical datasets pose a unique challenge for machine learning and data mining techniques to extract accurate, comprehensible and hidden knowledge from them. In this paper, we investigate the role of a biomedical dataset on the classification accuracy of an algorithm. To this end, we quantify the complexity of a biomedical dataset in terms of its missing values, imbalance ratio, noise and information gain. We have performed our experiments using six wellknown evolutionary rule learning algorithms -XCS, UCS, GAssist, cAnt-Miner, SLAVE and Ishibuchi -on 31 publicly available biomedical datasets. The results of our experiments and statistical analysis show that GAssist gives better classification results on majority of biomedical datasets among the compared schemes but cannot be categorized as the best classifier. Moreover, our analysis reveals that the nature of a biomedical dataset -not the selection of evolutionary algorithm -plays a major role in determining the classification accuracy of a dataset. We further show that noise is a dominating factor in determining the complexity of a dataset and it is inversely proportional to the classification accuracy of all evaluated algorithms. Towards the end, we provide researchers with a metaclassification model that can be used to determine the classification potential of a dataset on the basis of its complexity measures.

show abstract

“…The reason is that it evolves only those highly-rewarded classifiers of the match set in the correct set, which predict the same class as that of the training example [24]. In comparison, GAssist has serious problems in dealing with multi-class problems -specially when the when the number of output classes are more than 5.…”

Section: Role Of Multiple Classesmentioning

confidence: 99%

Performance evaluation of evolutionary algorithms in classification of biomedical datasets

Tanwani

Farooq

2009

Proceedings of the 11th Annual Conference Companion on Genetic and Evolutionary Computation Conference: Late Breaking Papers

View full text Add to dashboard Cite

Biomedical datasets pose a unique challenge for machine learning and data mining techniques to extract accurate, comprehensible and hidden knowledge from them. In this paper, we comprehensively investigate the role of a biomedical dataset on the classification accuracy of an algorithm. To this end, we quantify the complexity of a biomedical dataset in terms of its missing values, imbalance ratio, noise and information gain. We have performed our experiments using six well-known evolutionary rule learning algorithms: XCS, UCS, GAssist, cAnt-Miner, SLAVE and Ishibuchi, on 31 publicly available biomedical datasets. The results of our experiments show that GAssist gives better classification accuracy among the compared schemes. However, the nature of a biomedical dataset -not the selection of evolutionary algorithm -plays a major role in determining the classification accuracy of a dataset. We further show that noise is a dominating factor in determining the complexity of a dataset and it is inversely proportional to the classification accuracy of all the algorithms. The complexity of biomedical dataset will prove useful to researchers in evaluating the classification potential of their dataset for automatic knowledge extraction.

show abstract

Revisiting UCS: Description, Fitness Sharing, and Comparison with XCS

Cited by 32 publications

References 14 publications

Clustering-based initialization of Learning Classifier Systems

Clustering-based initialization of Learning Classifier Systems

Classification Potential vs. Classification Accuracy: A Comprehensive Study of Evolutionary Algorithms with Biomedical Datasets

Performance evaluation of evolutionary algorithms in classification of biomedical datasets

Contact Info

Product

Resources

About