Self-organizing feature maps for solving location-allocation problems with rectilinear distances
Computers and Operations Research
Review of the Self-Organizing Map (SOM) approach in water resources: Commentary
Environmental Modelling & Software
| Hi-index | 0.00 |
In the next decades, a significant increase is expected in the amounts of CCA-treated wood waste that annually need to be properly disposed. This waste should be recycled only after its remediation, so planning and optimisation of the remediation units location is of major importance. A location model for CCA-treated wood waste was implemented using Geographic Information Systems (ArcGIS 8.2), with geographic information, namely land use information and the results of a questionnaire sent to Portuguese wood preservation industries. Two different clustering methods (Self-Organizing Maps and K-means) were tested in different conditions to solve the multisource Weber problem using SOMToolbox for MATLAB. The solutions obtained with the data and with both clustering methods could be used to decide on the location of these plants. SOM provided more robust and reproducible results than K-means, with the disadvantage of longer computing times. The main advantage of K-means, compared to SOM, is the reduced computing time (considering an average of all the runs, the K-means computing time is half the SOM computing time) together with the fact that it allows to obtain the best solutions in the majority of the cases, in spite of bigger variances and more geographical dispersion.