A multiscalar tools set of "simulation outputs mining" :
In order to calibrate the model and to validate it, as well as to valorise the
different results of the simulations, two multiscalar tools of "simulation outputs mining"
have been developed.
The first tool is an interactive. It is dedicated to the calibration phase:
it is possible to visualize the state of each parameter or variable, at any time
t, of the simulation. A particular care has been put in editing windows for
defining the parameters that are a function of time (figure 1). The same tool
is used for a rough visualisation of the simulation. During the calibration phase,
it allows for instance to follow precisely the values of several attributes of one
city step by step, or to explore at a time t the different values of an attribute
among a set of cities. (figure 2). Some graphics representing global evolution are
also proposed for population, production...in order to control very quickly the order
of magnitude of some target variables.
The second tool is dedicated to the validation phase, during which the
main objectives are:
To test the coherence of the rules introduced in the model,
To evaluate the ability of the model to reproduce trajectories which
are realistic according to what is observed,
To get an insight in its sensitivity to initial conditions
and parameter variations.
A Control Widget for adjusting a parameter value / date
Snapshot of the Simpop2 Eurosim viewer
The potential outputs of the simulations represent a huge amount of data.
For instance in Eurosim, the output for one simulation concerns 5000 towns
and cities, 100 time steps, 13 functions and all interactions associated to
communications and exchanges between the cities. Therefore, the second tool
is less interactive. Its output presents a necessary systematic exploration
of the simulation results, focusing on three central dimensions :
The tool offers some systematic procedure of analysis of all the outputs of
a simulation. The principle is to "cut-out" the cube of information crossing
time, space and functions, using the complementarities of different
territorial filters, in order to get a quite complete idea of the different
states of the urban system during the successive periods.
Thus, rather than computing a single objective function, in the form of
an one-dimensional summary of the simulated values, this tool proposes a
multidimensional framework of evaluation of the simulation and integrates
a number of measurements on the structures themselves: hierarchical, spatial,
For each simulation, outputs are summarized and illustrated in a standard
report composed of three parts, corresponding to the three geographical
levels of interest : Some examples of the variety of outputs are given here.
The local/city level concerns the cities themselves. At this level, the report
focuses on the local trajectories associated to population, wealth, labour force,
variation rates...and spatial structures of exchange (maps of the market networks,
maps of exchanges flows, graphics of evolution of the size of the networks)
The different levels are also analysed through the hierarchical and spatial
dimensions of the urban system.