an example: neural network application and development

Networks have been built to determine the risk of heart attack in men and women. Some also determine the percentage risk of a heart attack proving fatal in a given period of time.

Data will be obtained representing the cases of thousands of people who have suffered heart attacks. The data might include age, own and family medical history, previous illnesses, blood pressure, cholesterol levels, stress factors, lifestyle, attitudes, support, diet, body mass index, smoking, alcohol consumption etc. Holding back a percentage of data for testing, the data is processed into a form suitable for acceptance by the network. A network is designed and trained to the results which might be survived heart attack, and/or terminal heart attack. Training is undertaken, through several iterations in which the network is improved until it tests satisfactorily on the unseen data. A user interface is developed which allows the user to predict the outcome for any new data added and to conduct a ‘what/if’ exercise (e.g. how are my chances improved if I stop smoking?).