IEEE Systems, Man and Cybernetics Magazine - October 2020 - 51

Tier 1: Poles

Cloud
Computing
iVAT

Tier 2: Buoys
Cyclone Hamish
Detection

Tier 3: Sensor Nodes

Base Station Tower

Pressure (Pa)

Heron Island (HI30)
1,020
1,010

1,000
10
20 30
Tem
pera
ture
(°C)

80
60
Humidity (%)

Figure 5. A two-tiered (tree) hierarchical network architecture of wireless sensor nodes deployed on Heron

faced by modern cities for better decision making. However,
interpretation of the massive amount of smart city-generated big data to create actionable knowledge is a challenging
task. Environmental sensors measuring luminosity, humidity, and temperature were deployed at Fitzroy Gardens and
Docklands Library to study the effects of canopy cover and
its impact on extrapolating said effects to enhance citizens'
interactions with the city's infrastructure.
To this effect, we clustered sensor data (luminosity, humidity, and temperature) from four sensors at
the Docklands Library utilizing sliding windows. We
resampled the data so that there was one measurement from all four sensors in 30-min intervals, resulting in a 12-dimensional feature vector. In total, we
have measurements spanning 72 days; with a window
size of two days, we have 96 samples. At each time
step, the inc-VAT/inc-iVAT includes a new data point
while the dec-VAT/dec-iVAT removes the oldest data
point. In the first time step, the inc-VAT/inc-iVAT
appends 96 samples to the MST, whereas in the last
step, the dec-VAT/dec-iVAT terminates when the sample size equals 2. Figure 7 shows the application of
inc-iVAT/dec-iVAT to the Docklands Library data to
detect anomalous samples.
Urban Mobility Patterns of Taxis in Singapore
Analyzing clusters is a fundamental challenge in trajectory
mining; however, existing trajectory clustering algorithms

Pressure (mbar)

Island of the GBR for the continuous monitoring of the reef. Using iVAT, we detected the passage of Cyclone
Hamish (2009).

Normal
Data Points

1,015
1,010
1,005
1,000
995
30
25
Tem
20
pera
ture
(°C)

Anomalous
Anomalous Data Points
Data Points (5 March)
(9 March)
60
80
100
Humidity (%)

Figure 6. Cyclone Hamish passed the Queensland

state of Australia during 4-14 March 2009. The iVAT
algorithm accurately clusters these events as two
anomalous clusters, corresponding to 5 and 9 March
2009, respectively.

are not appropriate for large numbers of trajectories in a
city's road network because of inadequate distance measurements between two trajectories. We utilized the GPS
traces of 15,061 taxis in Singapore (equaling 3.28 million
trajectories) gathered over a one-month period. To cluster
the origin and destination pairs of taxi rides, we used clusiVAT sampling and the density-based spatial clustering of
applications with noise [28] to provide useful insight into
urban hot spots, the usage of road networks, and crowd
movements. For large numbers of overlapping trajectories,
we used dynamic time warping (DTW) coupled with Dijkstra (TrajDTW) distance measurements [23], [29]. For
O c tob e r 2020

IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE

IEEE Systems, Man and Cybernetics Magazine - October 2020

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - October 2020