Cross-fertilization of conventional statistical techniques and Kohonen Maps, for clustering socio-economic data
With its unusual mixture between nearest neighbour classification and planar representation of the reference vectors, the Kohonen Map connectionist model has no analogous traditional method for unsupervised data analysis. This could be the reason why the information content of the connectionist approach may be nicely improved by applying traditional statistical techniques on top of the Kohonen Map. We illustrate this statement by performing a principal component analysis and a hierarchical clustering over the reference vectors of a Kohonen Map representation of a complex real-life dataset.
Bibliographic Reference: Article: Visula Intelligence for Finance, Investment and Information Management
Record Number: 199710770 / Last updated on: 1997-07-02
Original language: en
Available languages: en