Generating Fuzzy Partitions from Nominal and Numerical Attributes with Imprecise Values

In areas of Data Mining and Soft Computing is important the dis- cretization of numerical attributes because there are techniques that can not work with numerical domains or can get better results when working with discrete do- mains. The precision obtained with these techniques depends largely on the qual- ity of the discretization performed. Moreover, in many real-world applications, data from which the discretization is carried out, are imprecise. In this paper we address both problems by proposing an algorithm to obtain a fuzzy discretization of numerical attributes from input data that show imprecise values in both nu- merical and nominal attributes. To evaluate the proposed algorithm we analyze the results on a set of datasets from different real-world problems.