Future Performance Modeling in Athletism with Low Quality Data-based Genetic Fuzzy Systems

A fuzzy rule-based decision system for assisting coaches in the configuration of an athletics team is presented. The knowledge base of this system combines the experience of the trainer with genetically mined information from training sessions and competitions. The novelty of our approach comes from the fact that these sources of data have low quality: they include subjective perceptions of mistakes of the athletes, different measurements taken by different observers, and interval-valued attributes. We will use a possibilistic representation of these categories of information, in combination with an extension principle-based reasoning method, and show that the predictive power of a genetic fuzzy system which is based in these principles improves other systems that discard the vagueness of the training data.