Determining the maximum length of logical rules in a classifier and visual comparison of results

Supervised learning problems can be faced by using a wide variety of approaches supported in machine learning. In recent years there has been an increasing interest in using the evolutionary computation paradigm as the classifier search method, helping the technique of applied machine learning. In this context, the knowledge representation in form of logical rules has been one of the most accepted machine learning approaches, because of its level of expressiveness. This paper proposes an evolutionary framework for rule-based classifier induction and is based on the idea of sequential covering. We introduce genetic programming as the search method for classification-rules. From this approach, we have given results on subjects as maximum rule length, number of rules needed in a classifier and the rule intersection problem. The experiments developed on benchmark clinical data resulted in a methodology to follow in the learning method evaluation. Moreover, the results achieved compared to other methods have shown that our proposal can be very useful in data analysis and classification coming from the medical domain. • The method is based on genetic programming techniques to find rules holding each class in a dataset. • The method is approached to solve the problem of rule intersection from different classes. • The method states the maximum length of a rule to generalize.