Learning from incomplete training data with missing values and medical application

A neural-network-based classification system is constructed to handle incomplete data with missing attribute values, and applied to a medical diagnosis. In the authors' approach, unknown values are represented by intervals. Therefore incomplete data with missing attribute values are transformed into interval data. A learning algorithm for multiclass classification problems of interval input vectors is derived. The proposed approach is applied to the medical diagnosis of hepatic diseases, and its performance is compared with that of a rule-based fuzzy classification system.