Clustering Contextual Neural Gas: A New Approach for Spatial Planning and Analysis Tasks - Computational Approaches for Urban Environments - page 89

rank ordering to form a topology, it is also possible to remove the first neuron
temporarily from the set of neurons, determine a new rank order using the CNG's
ordering scheme, and then connect the first neuron of the resulting rank order with
the previously found first neuron. How this strategy performs in comparison to the
one used in this study is unclear and deserves further research. Additionally, CHL is
sensitive to noisy data and outliers (Aupetit 2005 ). Using alternative algorithms for
topology learning bears potential to improve the results.
This study uses the MLMO algorithm to cluster the CNG's topology. The MLMO
algorithm uses a greedy heuristic to optimize the modularity score of the graph.
Although the algorithm has been shown to generally perform very well, it lacks
accuracy, like any greedy clustering method (Fortunato 2010 ). In principle, any
other graph clustering algorithm can be applied within the graph clustering step
of the method.
The proposed method combines different methods from different but related
disciplines for clustering spatial data. As scientific research for each of these
disciplines is going to continue, it can be expected that more powerful methods will
be developed. Utilizing these methods has the potential to further increase the value
of the proposed method. In particular, improving the CNG algorithm with regard to
convergence and parametrization seems worth pursuing.
Finally, the presented method is rather technical and difficult to understand and
apply by nonexperts. In order to be of real practical value for spatial planners and
policy makers, it is necessary to integrate the method into a combined software
toolkit which provides powerful analytical and visual means in order to validate the
results and which is also easy to use.
References
Amedeo D (1969) An optimization approach to the identification of a system of regions. Pap Reg
Sci 23(1):25-44
Arribas-Bel D, Schmidt CR (2013) Self-organizing maps and the US urban spatial structure.
Environ Plan B Plan Des 40(2):362-371
Aupetit M (2005) Learning topology with the generative gaussian graph and the EM algorithm. In:
Advances in neural information processing systems, Vancouver, pp 83-90
Bação F, Lobo V, Painho M (2005) The self-organizing map, the geo-som, and relevant variants
for geosciences. Comput Geosci 31:155-163
Bailey T, Gatrell A (1995) Interactive spatial data analysis. Longman Scientific & Technical,
Harlow Essex
Blondel VD, Guillaume JL, Lambiotte R, Lefebvre E (2008) Fast unfolding of communities in
large networks. J Stat Mech Theory Exp 2008(10):P10008
Brandes U, Delling D, Gaertler M, Gorke R, Hoefer M, Nikoloski Z, Wagner D (2008) On
modularity clustering. IEEE Trans Knowl Data Eng 20(2):172-188
Costa J, De Andrade Netto M (1999) Cluster analysis using self-organizing maps and image
processing techniques. In: IEEE international conference on systems, man, and cybernetics
1999. IEEE SMC '99 conference proceedings, Tokyo, vol 5, pp 367-372
Costa J, Oliveira R (2007) Cluster analysis using growing neural gas and graph partitioning. In:
International joint conference on neural networks, IJCNN 2007, Orlando, pp 3051-3056