Evaluation of classification results with the disjoint classification program SOLOMON

Practical application of pattern recognition techniques is often limited because of the way in which variables are selected and the need for supplementation of missing data. The software package comprising the MICA transformation technique and the SOLOMON classification program is extended to include an algorithm for estimating and supplementing missing data. The performance of this package is compared with that of known classification techniques on simulated data sets. The supplementation algorithm is tested with a data set for hyperlipoproteinemia patients.