A neural network to predict attendance of paediatric patients at outpatient clinics

Abstract

The development of a feed-forward neural network to predict attendance and non-attendance at a hospital outpatient clinic is described. Particular emphasis is given to the post-processing applied to the network output, and to the choice of information presented to the network. The effect of individual inputs on the performance of the network is examined, and two methods of dealing with the problem of ambiguous input data are explored. It is shown that good results (approximately 90% accuracy) are achieved with a feedforward network — having 15 input nodes, a single hidden layer of five nodes and one output node with supervised backpropagation — even where data is limited, and that an understanding of the problem can lead to the modification of standard procedures and improved network performance.