The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #9810

Title:Investigating Genetic Algorithms for Improving Diagnostic Accuracy in Malarial Risk Assessment
Date: 1998
Abstract:The use of personal computers to aid medical diagnosis continues to increase in the developing world. Recent research into malaria diagnosis by the World Health Organisation recommends adding estimation of anaemia levels to the traditional malariometic survey. The trial of a malaria risk assessment system might lead to further development of a robust and reliable prototype.Conventional case-based reasoning is shown to perform well on malarial risk assessment and its accuracy can be further improved with the use of feature weights. A genetic algorithm is used to select suitable feature weights, and its performance compared with that of a population based hill climber.Whilst the genetic algorithm is found to produce higher accuracy than the hill climber on average, both techniques result in a significantly higher accuracy than without the use of feature weights. More research is required to see if these promising results will transfer to other, larger malarial case-bases.

[Search These Pages] [DAI Home Page] [Comment]