3. Neural Network Definition for Modelling in Spatial Dimension
As part of travel forecast model definition, spatial and temporal dimensions are separately
analyzed. Spatial dimension is devoted to represent the basic relations between land use patterns
and transportation system affecting travel demand in a static stage of time. In the sequence, it is
described NN formulation for spatial dimension.
Consider an urban area, which is divided in Traffic Zones (TZ). Trips
(T
ij) between T
Z pairs are
verified, where i and j denote the origin and destination, respectively. Land use is expressed by
occupied area of each pattern in each TZ,
which is assigned to RLU
TZ (Residential Land use)
CLU
TZ (Commercial Land use) SLU
TZ (Service Land use). Transportation system is characterized
by the extension of each mode in each zone, which is associated to RTS
TZ
(Road Transportation
System), BTS
TZ (Bus Transportation System),
STS
TZ (Subway Transportation System) and TTS
TZ
(Train Transportation System).
Figure 1: RS and GIS and NN integration for Travel Forecast Problem
To be processed as a feedforward Multilayer Perceptron (see Wasserman, 1989), these
characteristics are transformed in input (X) and output (Y) vectors. Then, it is defined that X =
(RLU
i, CLU
i,
SLU
i, RTS
i, BTS
i,
STS
i, TTS
i, RLU
j, CLU
j,
SLU
j, RTS
j,
BTS
j, STS
j, TTS
j) and Y=
(T
ij). Physically, it means that trips between two zones (i and j) are expressed in NN by the Land
use and transportation systems indicators for each zone (origin and destination). As a typical NN,
input, hidden and output layers compose it. Hidden layers (H) are composed by are hidden
neuron units H
1, H
2, ... H
m connect input and output, expressed by X and Y, respectively, as
shown by Figure 2.
Figure 2: Typical NN structure
The output of the neurons are calculated by:
where f
a and f
b are activation functions that are suggested, in this work, to be logistic bipolar
and identity function, respectively. W and V weight matrixes are obtained by applying a
backpropagation based training algorithm (Wasserman, 1989).
4. Preliminary Tests
Tests were conducted in Boston Metropolitan Area (Massachusetts State – USA) (see Figure 3a),
which covers about 1400 square miles (3580 square kilometers) where nearly three million
people live. It was selected a study area in Boston South area, near to Boston Medical Center
that involves seventeen TZ`s as shown in Figure 3b. Study area is mainly occupied by residential
land use and it is located very close to downtown.
RS and transportation system data were obtained from MassGIS database. In this database, data
was projected to Massachusetts State plane Mainland Zone coordinate system, Datum NAD83
in meters. It was used black and white digital orthophotos produced in 1992 in 1:5000 scale. Bus
route maps from Metropolitan Boston Transportation Authority (MBTA) were incorporated to
Transportation data. Finally, Central Transportation Planning Staff (CTPS) provided access to
travel data related to 1990 survey that involves all purposes trips as well as TZ definition.
In the sequence, GIS database construction activities are described and results of the
experiments are introduced.
4.1 GIS database
Geoconcept GIS software was used in these tests. First, land use patterns were obtained
following United States Geological Service (USGS) classification system (Avery and Berlin,
1990) and Taco’s methodology (Taco et al, 1999). It was analyzed up to Level II (Residential,
Commercial and Services) according to model’s requirements (see item 3). Figure 3c presents
the land use patterns for the study area. Next, transportation system was digitalized in the GIS
database. Figure 3d shows road, bus, subway and train systems for study area.
