Information Technology Reference
In-Depth Information
•
Flow Category 1-4: Type of vehicle based on their length. Categories 1 and 2
represent light vehicles, while categories 3 and 4 represent heavy good vehicles.
•
Average Headway: Arithmetic mean of the distance between vehicles measured
with a precision of 0.1s.
•
Occupancy: Ratio of the total detection time to the maximum detection time, ex-
pressed as a percentage, where the detection time refers to the amount of time in
which the sensor detects the presence of a vehicle.
Accident Data.
The WTTC also keeps a record of incidents such as accidents, con-
gestion, road works, etc. We use the accident dataset, which contains information that
specifies the location, time and type of every accident.
3.2
Architecture of the Experimental Setup
The main requirement of an automated traffic accident forecasting system is to identi-
fy when flow data indicates a likelihood of causing an accident. In order to study the
feasibility of using CBR for this purpose, the capability of the CBR system to diffe-
rentiate between historical flow data that has led to an accident and historical flow
data that was not followed by an accident is investigated.
Target case
• Time interval [
t
m
, t
m-1
,…, t
1
]
•Distance [
d
n
, d
n-1
, …, d
1
]
•Outcome (C
T
)
Case Base
• Time interval [
t
m
, t
m-1
,…, t
1
]
•Distance [
d
n
, d
n-1
, …, d
1
]
•Outcome (C
C
)
Retrieve most similar case
Similarity calculation
Unsuccessful
retrieval
Successful
retrieval
Is Outcome(C
T
) = Outcome(C
C
)?
NO
YES
Fig. 2.
Schematic diagram showing steps involved to evaluate the differentiation capability of
the CBR system
Fig. 2 shows a schematic overview of the CBR setup to determine if the system is
capable of differentiating between accidents and non-accidents. The current flow data
to be analysed for its accident risk constitutes the target case. The flow data is consi-
dered over a pre-specified interval of time [
t
m
, t
1
] and distance [
d
n
,d
1
], where
m
and
n
refer to the point or index of a sensor with respect to time or distance, respectively.
