Knowledge discovering from clinical data based on classification tasks solving.

Abstract

After the abundant literature on normative decision analysis centered on Bayes' rule, the artificial intelligence approach left aside the proposed formalism in favor of symbolic approach. Thus, diagnosis was identified as a dynamic cognitive process, characterized by the search for evidence to test a given hypothesis where heuristic thinking plays a significant role. Approach presented in this study is knowledge discovering system to test a given (manually or automatically) hypothesis, based on three mathematical models of classification: model of classes stability, model of linear envelope, model of multilayer neural network. All the models characterizes as quantitative methods, and works as cognitive tools for new laws searching.