The urban sprawl dynamics: does a neural network understand the spatial logic better than a cellular automata?



unit vector, but also its neighbourhood according to a decreasing function (fig 1b).

SOM NN on square grids of 9,16, 25 nodes have been trained to group the data. The
more explaining output was obtained with a 4x4 nodes grid; this resulted the best in
order to show each group sufficiently different from each other.


c12                                    c13                                    c14


c21                                    c22                                    c23                                    c24



c33



CR80 Cell, residential, 1980

CP80 Cell, productive, 1980

CC80 Cell, commercial, 1980

RD Road distance

NR80 Neighbourhood, residential, 1980

NP80 Neighbourhood, productive, 1980

NC80 Neighbourhood, commercial, 1980


CR94 Cell, residential, 1994

CP94 Cell, productive, 1994

CC94 Cell, commercial, 1994


c43


I Codebooks range, values over the mean

Codebooks range, values under the mean

—— Codebook

Figure 2 - The codebooks

The spatial analysis carried out by SOM has been displayed by:

cluster profiles and their codebook;

charts with colour hatched plot of the zones, based on output units assignment.

In figure 2 all the codebooks, as prototypical profiles of each cluster, shows the most
relevant features in land use dynamics. On the x-axis are the variables, on the y-axis
their activation level. On the figure is charted the envelope of the records assigned to
each single cluster and, in yellow line, the
codebook.

The colour map (Fig.3) shows the spatial organisation of the classes; it is crucial to
know if cells belonging to the same class are also spatially clustered, or if similar



More intriguing information

1. The Role of Trait Emotional Intelligence (El) in the Workplace.
2. AN EMPIRICAL INVESTIGATION OF THE PRODUCTION EFFECTS OF ADOPTING GM SEED TECHNOLOGY: THE CASE OF FARMERS IN ARGENTINA
3. Globalization and the benefits of trade
4. Top-Down Mass Analysis of Protein Tyrosine Nitration: Comparison of Electron Capture Dissociation with “Slow-Heating” Tandem Mass Spectrometry Methods
5. Educational Inequalities Among School Leavers in Ireland 1979-1994
6. Dendritic Inhibition Enhances Neural Coding Properties
7. Consumption Behaviour in Zambia: The Link to Poverty Alleviation?
8. The name is absent
9. The Prohibition of the Proposed Springer-ProSiebenSat.1-Merger: How much Economics in German Merger Control?
10. The name is absent