Information Technology Reference
In-Depth Information
4.
Spatial attributes
These attributes describe the variation in the attributes of traffic flow data over dis-
tance
L.
The spatial variations are determined by calculating the net increase or de-
crease between the starting point of the stretch of road considered and the end (or
current) point of the traffic flow attributes. They include:
•
Variation in average velocity over all lanes
V
L,s
,
,
,
(3)
where
s
L,Start
and
s
L,End
are the average speed of vehicles over all lanes found at the
starting and end point, respectively, of the distance
L
considered.
•
Variation in average flow over all lanes
V
L,f
,
,
,
(4)
where
f
L,Start
and
f
L,End
are the average number of vehicles over all lanes found at the
starting and end point respectively of the distance
L
considered.
3.4
Retrieval and Similarity Measure
The retrieval mechanism constitutes the inference engine of the CBR system. Given a
target case, the system identifies the most similar case in the case base to the target case
and retrieves it. The main component of the retrieval mechanism is the similarity meas-
ure, which calculates the similarity between the target case and the cases in the case
base. Since CBR systems are based on the notion of similar cases having similar solu-
tions, the definition of similarity is crucial. The CBR system depends on a good retriev-
al engine that is capable of retrieving cases whose solution is relevant to the target case.
The similarity in our CBR system is computed using the
K
-nearest neighbour algo-
rithm (KNN), where
K
refers to the number of neighbours. The KNN algorithm, tradi-
tionally used in classification and pattern recognition problems to assign objects to
classes, has been widely used in CBR as well, owing to its ease of implementation
and the fact that it does not make any assumptions about the distribution of the under-
lying data. In KNN classification, an object is classified by assigning it to the known
class of its nearest examples (or neighbours) in the solution space. The solution space
in CBR can be viewed as a collection of clusters, where each cluster contains similar
solutions [18]. The nearest neighbour is found using a distance metric defined for
each type of attribute. Currently, we consider four combinations of attribute similari-
ties in the similarity measure
•
Time - Day Category Similarity,
S
TD
:
If the target case and a case from the case base fall into the same time-day category,
the similarity
S
TD
is '1', else it is '0'.
