homogenous. As such, urban transport planners would have to
structure more dynamic policies that take into considerations
the heterogeneous passenger travel patterns.
For example, the Land Transport Authority of Singapore
had decided to allow public train commuters to travel free on
weekdays if the commuters exit 16 stations in central city area
before 7.45AM [17]. The objective of this initiative is to get
public train commuters to travel earlier so as to ease off the
huge surge in passenger volume in late evening. Based on the
findings in this research, an improvement to this initiative
could be allowing commuters to travel free if they could enter
residential area stations before their peak hour. This
improvement will allow more certainty in easing off
passenger volumes of the origin stations and prevent building
huge crowd at the destination at 7.45AM.
A. Research Limitations and Future Works
This research had demonstrated the usefulness of
time-series data mining techniques for knowledge discovered
on Singapore‘s public train passenger travel patterns.
However, there are dimensions that this research did not cover
and would be the potential areas for future works. Some of
these areas include:
Exit Timing. This research was done based on the entry
timestamp. It would complete the analysis if another
time-series data mining were done based on the exit
timestamp of passengers exiting the train stations.
Gravity Model of Migration As seen in each of the
cluster analysis and insights interpretation, it would be helpful
if we could ascertain if indeed the train stations are situation in
residential, commercial office or retail areas. This will help us
to explain the public train passengers‘ travel patterns in
greater details.
Predictive Analytics. Predictive analytics can be done
using the results of this research to predict how the passengers
would behave when the public train extension works for 2020
are completed.
VIII. CONCLUSION
With the application of time-series data mining techniques
and sensing data in transportation studies, urban transport
planners and analysts will be able to analyze the passenger
travel patterns faster and gain greater insights beyond what
could be provided by conventional statistical analysis or
traditional data mining techniques. There are also a number of
future works that could be done to generate greater insights
and knowledge discovery. The time-series data mining
framework proposed in this research is also extensible to
study other transport modes such as buses and taxis, and for
other cities‘ transportation networks too.
ACKNOWLEDGMENT
We would like to thank the Land Transport Authority (LTA)
of Singapore for sharing with us the MRT dataset. This
research is supported by the Singapore National Research
Foundation under its International Research Centre @
Singapore Funding Initiative and administered by the IDM
Programme Office, Media Development Authority (MDA).