Hypertension and socio-demography: using neural networks to predict high blood pressure from responses to the English Longitudinal Study of Ageing

Abstract — The present study explores two classes of neural computing algorithms (i.e. Support Vector Machines (SVMs); and Multilayer Perceptrons (MLPs) trained with backpropagation) with regard to their accuracy in predicting self-reported hypertension based on survey data from the English Longitudinal Study of Ageing. The findings suggest that the performance of both algorithms declines when they are trained using inputs derived by applying Self-Organising Maps (SOMs) on the original inputs. Against our expectation, the findings suggest that SVMs outperform MLPs; however, only by a very small margin. Implications for the domain of social- science analytics are briefly considered.