Contextual Approach to Data Discretization

This paper presents a new discretization algorithm that takes into account the behavior of associated variables. Indeed, in the context of association rules extraction, for example, the goal is to find interconnected data. Thus, instead of computing numeric variables independently we choose to compute them in their context, i.e. in association with the rest of the variables to consider. The proposed approach is based on the joint use of statistical constraints (objective measures) that are in charge of determining the real significance of the relationships between variables and human constraints (subjective measures) defined by the domain expert and concerning thresholds determination.