Denny,
Visualizations of Cluster Changes by Comparing
Self-Organizing Maps.
Master's Minor Thesis, School of Computer Science and Software Engineering,
Monash University, 900 Dandenong Road Caulfield, Victoria 3145, Australia, July 2004.
Denny,
Visualizations of Cluster Changes by Comparing
Self-Organizing Maps.
Master's Minor Thesis, School of Computer Science and Software Engineering,
Monash University, 900 Dandenong Road Caulfield, Victoria 3145, Australia, July 2004.
In today's fast-moving era, knowledge of change is important for an organization in order for it to quickly adapt its strategies. For instance, say an organization has devised marketing strategies based on last year's clustering results of their customers. It would be important for the organization to know whether this year's clustering results are different compared to last year's results, in order to review and perhaps revise their marketing strategies. Therefore, we propose a SOM training method and SOM-based visualization techniques that are capable of explaining the clustering result of the second dataset in terms of the clustering result of the first dataset by using colour and position linking. The proposed solution was implemented using the SOM Toolbox and also the MATLAB. Performance tests with synthetic datasets and real-life datasets using the World Development Indicator data published by the World Bank show that the proposed visualization methods can show structural changes which are: changes in distributions, centroid movements, new clusters, and missing clusters. Furthermore, the proposed visualization methods can also show migration of the cluster members on the map.