To Handle Real Valued Input in XCS: Using Fuzzy Hyper-trapezoidal Membership in Classifier Condition

Learning classifier systems (LCSs) are evolutionary learning mechanisms that combine genetic algorithms (GAs) with the power of the reinforcement learning paradigm. XCS, eXtended Classifier System, is currently considered as state of the art learning classifier systems due to its effectiveness in data analysis and its success in applying to varieties of learning problems. Generalization and the size of evolved population in XCS is one of the most challenging issues in XCS. This paper suggests that rule representation in XCS is not matured to the point where the condition parts of classifiers are covering the problem space in an effective manner with minimum redundancy. The key idea in the proposed system, named XCS-HT, is to use a novel representation scheme based on presenting a subspace of problem space with two kinds of regions named certain and vague regions. Using such mechanism significantly boils the number of evolved XCS-HT’s classifiers down, while changing the performance only marginally. The experimental results show that XCS-HT has a better ability to solve problems having indistinguishable class boundaries comparing to XCSR, a common extension of XCS standing for handling real valued data.