A consumer health decision support system (CDSS) is being developed at the South African Medical Research Council (MRC). It is a software program intended to help members of the public decide when they may be at risk of some common but serious illnesses like tuberculosis and hypertension. It would be ideal for a public health kiosk or e-health programs of the government. The program has been built as an expert system. Its knowledge base consists of rules which are used in assessing the risk of illness. The rules were given by medical experts who took part in the development of the CDSS. The study proposes a method for the evaluation of the rule base of the CDSS using FCA methods. It is important to evaluate the knowledge base of an expert system, because if its knowledge base is of broad scope and is accurate then it can be expected that the expert system will be good at giving advice and hence potentially useful. FCA is a mathematical framework which can be used to investigate causal relations in data. The study explored its utility in the evaluation of the CDSS knowledge base. FCA implications and the FCA formulation of the JSM method were two FCA methods that were selected. The FCA methods were used to generate rules from actual patient data, and these were compared to the rules initially given by the experts. The motivation to use FCA data analysis as well as experts’ knowledge in the development of the CDSS program is that FCA data analysis may discover some things that the experts may have overlooked. Or at least the experts can review their expertise against actual field data which has been analysed by FCA methods. A system like the CDSS cannot be built using FCA data analysis techniques only, involvement of experts is very important. The two FCA methods were chosen so as to compare their results, and it was also thought that they may perhaps complement each other. Preliminarily it was found that FCA implications and the FCA formulation of the JSM method can be used in the evaluation of the rule base of the CDSS. / Dissertation (MSc (Computer Science))--University of Pretoria, 2008. / Computer Science / unrestricted
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:up/oai:repository.up.ac.za:2263/24344 |
| Date | 05 May 2008 |
| Creators | Horner, Vincent Zion |
| Contributors | Prof D Kourie, Dr. S Obiedkov, vincent.horner@mrc.ac.za |
| Source Sets | South African National ETD Portal |
| Detected Language | English |
| Type | Dissertation |
| Rights | © University of Pretoria 2007E829 / |
Page generated in 0.0023 seconds