Uncertainty based approach for symbolic classification of numeric variables in intensive care units

In intensive care units (ICUs), certain parameters must be interpreted taking into account the intrinsic characteristics of each patient or the peculiarities of the clinical case under consideration. Similar temporal evolutions of some parameters in different patients could have different interpretations. Artificial intelligence techniques can aid in resolving this problem through the construction of expert systems (ES). These systems are capable of performing contextual evaluations of the parameters typically monitored in ICUs. This contextual evaluation is usually carried out using symbolic elements. Thus, the symbolic processing of numeric data is an important task to perform. In any event, the assignment of semantic labels to numeric values is always an uncertain and arbitrary process. This suggests the convenience of defining and implementing representation schemes capable of dealing with uncertain knowledge. This paper presents a model for the symbolic processing of numeric variables in which the uncertainty associated with the assignment of literals appears spontaneously. The categorical approach for the symbolic classification of numeric values is a particular case of the proposed model.