Application of multiple self-organizing maps for classification of soil samples in Thailand according to their geographic origins

Multiple self-organizing maps (SOMs) were applied to classify soil samples according to their geographic origins. The soil physical and chemical parameters, including textures, pH, and chemical nutrients, were analyzed and used for establishing the chemometric models. To determine the optimum size and arrangement of the maps, we adapted a growing self-organizing map algorithm. To evaluate the reliability of the models, we calculated statistic indices based on the majority vote including percentage predictive ability, percentage model stability, and percentage correctly classified using a bootstrap methodology. For means of comparison, we also used linear discriminant analysis, quadratic discriminant analysis, partial least squares-discriminant analysis, soft independent modeling of class analogy, counter propagation network, supervised Kohonen network, and k-nearest neighbors. In comparison to a single SOM, multiple SOMs clearly provided better classification results. The extension of multiple SOMs also led to the best discrimination of the soil origins.