Neural-network-based diagnosis systems for incomplete data with missing inputs

The aim of this paper is to propose classification methods for incomplete data with missing inputs in neural-network-based diagnosis systems. In this paper, such incomplete data are treated as intervals by representing each missing input by the range of its possible values. We propose four definitions of inequality between intervals to classify new interval input vectors by neural networks. The performance of neural-network-based diagnosis systems with the proposed four definitions is examined by computer simulations on a diagnosis problem of hepatic diseases.<>