Automated Extraction of Decision Rules from Medical Databases - A Rough Sets Approach

One of the three goals of this work is to research automated knowledge acquisition process from table-organized data by using the Rough set theory. This theory proved to be an excellent mathematical tool for the task of automated extraction of rule sets from table-organized data. The second goal is to employ a system for automated theorem proving -LogPro, based on ordered linear resolution with marked literals. This approach has shown better performances that the one based on linear resolution with selection function for definite clauses and negation as failure, e.g. Prolog. The rule sets produced by rough sets technique, in form of Prolog-like clauses are generated from real life medical table-organized data including 36 parameters, one of them being protein hormone Leptin. The third goal of this work is to get new insight into phenomena of Leptin levels while interplaying with other risk factors in obesity, especially with Trunk Fat. This is done by employing generated rule sets in LogPro.