Classification of benchmark wine data by LVQ

Abstract

We have been comparing LVQ(Learning Vector Quantization) and multivariate discriminant analysis using model data and data used as a benchmark. When the data is normally distributed, there is almost no difference between the results of LVQ and multivariate discriminant analysis method. A multivariate discriminant analysis is a linear discrimination method. On the other hand, LVQ is a nonlinear discrimination one. When data is largely out of the normal distribution, good recognition result is obtained for LVQ. In this study, we compared LVQ with multivariate discriminant analysis for wine data. The property and usefulness of LVQ is shown in this report.