First study of the behaviour of genetic fuzzy classifier based on low quality data respect to the preprocessing of low quality imbalanced datasets

There are real-world dataset where we can found classes with a very different percentage of patterns between them, that is to say we have classes represented by many examples (high percentage of patterns) and classes represented by few examples (low percentage of patterns). These kind of datasets receive the name of “imbalanced datasets”. In the field of classification problems the imbalanced dataset are a focus of study both in preprocessing mechanisms and in classification systems. In this paper we study the behaviour of genetic fuzzy system (GFS) respect to imbalanced datasets, where this GFS is able to support low quality data. We will analyse the different preprocessing mechanisms of imbalanced datasets and will show the necessity of extending theses preprocessing mechanisms a “low quality imbalanced datasets”. In addition, we include a comprehensive description of the new algorithm to able to preprocessing low quality imbalanced datasets. Severals realworld, low quality imbalanced datasets, are used to evaluate the results obtained with the GFS after using the new algorithm proposed in this paper.