A Baseline Learning Genetic Fuzzy Classifier Based on Low Quality Data

Obtaining fuzzy rules from low quality data is a topic that has been recently formalized. This paper contains the first application of these principles to classification problems. We intend that the classifier proposed here serves as a baseline for future developments in the field. For that reason, we have extended a simple crisp genetic fuzzy classifier to imprecise data, paying special attention to the computational details. In particular, we will discuss some issues about the fuzzy-valued fitness function that is used in our formalism. A synthetic problem, plus two real-world datasets of low and medium complexities are also proposed, and used to evaluate the algorithm.