Knowledge discovery with clustering based on rules by states: a water treatment application

Abstract

This work presents advances in the design of a hybrid methodology that combines artificial intelligence and statistical tools to induce a model of explicit knowledge in relation to the dynamics of a wastewater treatment plant. The methodology contributes to problem solving under the paradigm of knowledge discovery from data in which the pre-process, the automatic interpretation of results and the explicit production of knowledge play a role as important as the analysis itself. The data mining step is performed using clustering based on rules by states, which integrates the knowledge discovered separately at each step of the process into a single model of global operation of the phenomenon. This provides a more accurate model for the dynamics of the system than one obtained by analyzing the whole dataset with all the steps taken together.