bul 2004
aÍ-
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
USER MODEL BUILDING
To build a personalised tourist support system, it is
important for the software agent to know what an
individual tourist likes. User preferences has been
widely studied and applied in e-commerce, where
vector representation is mostly used to represent
both the products to be recommended and the user's
preferences of these products (Lee et al., 2002;
Rogers ef al., 1999; Yuan and Tsao, 2003).
In the proposed tourist route advice system, the
products to be recommended are the tourist routes.
Each route is composed of a series of connected
road segments. The construction of the route is an
iterative process by selecting the most attractive
road segment at each step based on the tourist's
personal interests. This personal interests is stored
as the following vector for each tourist:
(Wi, Wy, Wy... WW.)
where /<= i <=m, m is the total number of
attributes characterising the tourist resources. Table
I gives an example of this vector for tourist X, in
which value 0 indicates that he is not at all
interested in the corresponding attribute and value
100 means the highest priority.
Table 1 Attributes of tourist resources in a study
area
Attributes Attributes Tourist
category weights on
attributes
Landscape | forest 90
heath 60
meadow 30
water area 100
agriculture field | 0
Roadtype | main road 0
local road 30
cycle path 70
foot path 100
Points of church 70
interest wind-mill 100
(on Farm house 70
restaurant 0
children- 0
D = playground
The next step in building the user model is to
calculate the so-called "attraction value" for cach
road segment. The attraction value is calculated
based on both the tourist resources available along
the road segments and tourist's personal interests.
The classification and the assessment of the tourist
resources is an important data and is usually carried
out by tourist experts. For example, the attributes
identified in Table 1 is typically related to
recreational walk or cycling. In addition, the experts
assess the attraction values of each individual tourist
resources based on the general tourist evaluation
and the (special) characteristics of the tourist
features. The assessment is also expressed as a
numeric value between 0 and 100, with 100
indicating the highest tourist attraction value. The
final attraction value of a particular tourist feature is
the multiplication of the tourist's interests and the
expert's assessed value. For example, if an area of
forest is assessed by experts as having attraction
value of 70 and a tourist gives a value of 90 to the
forest features in general, the final attraction value
of this forest to this tourist would be 70 x 90 = 63.
DATA PREPARATION
Before constructing a route for a tourist, the
database has to be prepared. In a GIS, a road
network is modelled as geo-referenced network and,
for the efficiency of computation, the geo-
referenced network can be abstracted as a planar
graph that preserves the structure of the geo-
referenced network. As explained later, the route
construction algorithm is entirely based on the road
network. Therefore, we have to link the tourist
attraction values calculated before with the road
segments.
Calculation of the attraction values for the attributes
under the landscape category in Table 1 is
complicated. They are related to the scenery
attractions along the tourist routes and are
represented as area features in the GIS database.
However, the algorithm applied in our tourist route
advice system considers only road network.
Therefore, a link must be built between the
landscape area features and the road network
elements, i.e., edges and nodes. This link is al
ready implicitly recorded in the GIS database
because all features in a GIS database are geo-
referenced and the topological relationships of these
features is also implicitly stored. The buffer
operation provided in a GIS tool is used here to
reveal this relationship:
— A buffer polygon is generated along each road
segment with a distance, for example 15
meters, along both sides of a road segment(see
figure 2). The value of this buffer distance can
be adjusted depending on the situations, such
as the scenery visibility along tourist routes.
— The percentage of the area of each landscape
feature within the total buffered area is then
calculated. These percentage values represent
the availability of each landscape feature on the
corresponding road segment.
To derive the attraction values of each road segment
for each tourist, we multiply the percentage values
with the attraction values calculated in the previous
step. The sum of these values represents the total
attraction value of the landscape on each road
segment.
