Abstract
Neural networks, a huge parallel interconnected network of simple elements performs adaptive learning from training data of a particular problem. In this study, a neural network design for the diagnosis of various diseases was done. The diseases incorporated in the study were; chicken pox, measles, dengue and flu. In order to achieve this, a methodology with back propagation model neural network model was used. Using this approach, feasible and infeasible design regions and boundaries that surround the sample space were derived from the sample data collected from medical centres in Kuala Lumpur, Malaysia. The data collection method used was questionnaire. The inputs to the neural network were disease symptoms like high temperature, headache, nausea, vomiting, rash, joint pain, muscle pain, bleeding, loss of appetite, diarrhoea, cough, sore throat, abdominal pain and runny nose. The study sample consisted of 32 patients with chicken pox, 34 with measles, 50 with dengue and 18 with flu. This neural network was trained and tested. A 94% achievement for correct disease diagnosis was attained. Classification is a vital tool in disease diagnosis decision support. This is achieved by using the feed forward back propagation network to distinguish between the various disease symptoms and associate them with the corresponding disease. This neural network approach to medical diagnosis signifies the potential of the neural network to learn patterns corresponding to symptoms