Towards Automatic Generation of Conceptual Interpretation of Clustering

In this paper the Methodology of conceptual characterization by embedded conditioning CCEC, oriented to the automatic generation of conceptual descriptions of classifications that can support later decision-making is presented, as well as its application to the interpretation of previously identified classes characterizing the different situations on a WasteWater Treatment Plant (WWTP). The particularity of the method is that it provides an interpretation of a partition previously obtained on an ill-structured domain, starting from a hierarchical clustering. The methodology uses some statistical tools (as the boxplot multiple, introduced by Tukey, which in our context behave as a powerful tool for numeric variables) together with some machine learning methods, to learn the structure of the data; this allows extracting useful information (using the concept of characterizing variable) for the automatic generation of a set of useful rules for later identification of classes. In this paper the usefulness of CCEC for building domain theories as models supporting later decision-making is addressed.