1 January - March 2015
Supervised Self Organizing Maps for Exploratory Data
Analysis of Running Waters Based on Physicochemical
Parameters: a Case Study in Chiang Mai, Thailand
Sila Kittiwachana* and Kate Grudpan
Department of Chemistry, Faculty of Science, Chiang Mai University, Chiang Mai, 50200 Thailand
Center of Excellence for Innovation in Analytical Science and Technology, Chiang Mai University,
Chiang Mai, 50200 Thailand
* Corresponding Author: email@example.com
This report demonstrated the use of a supervised self organizing map (SOM) for
exploratory analysis of running waters based on their chemical criteria. Water samples
from 10 different sites, representing 4 different water types – streams, a river, an irrigation
canal and a sewage canal – were collected from some areas in Chiang Mai, Thailand, during
8-month period from May to December and analyzed for 16 physicochemical parameters.
The samples were categorised into 8 classes (the 8 months from May to December) and
10 classes (the 10 sampling sites). This information was incorporated into the modeling
using a supervised SOM methodology. The results were visualized using supervised colour
shading and a unified distance matrix (U-matrix). The supervised SOM
improved the correlation among the samples within group. It was possible to reveal the
water sample clusters, either when organized according to the sampling times or sites.
Moreover, all of the variation could be used for the analysis, eliminating the need to choose
the specifi dimensions or the number of principal components (PCs).
Keywords: exploratory data analysis, supervised self organizing map (SOM), artifiial
neural networks (ANNs), principal component analysis (PCA), water analysis.