Research
Projects during PhD Study
|
Figure 1: Mobility
Profiler Framework
Figure 2: A
Representation of Mobility Profile |
Mobility
Profiler: A
Framework For Discovering Mobility Profiles of Cell Phone Users Problem: The location logs that can
be collected from cell phones are low level data units (either collected in
cellular or GPS enabled environment). These preliminary data units make
difficult to access mobility profiles of the cell phone users for several
applications. To make mobility data more accessible to cell phone
applications, higher level data abstractions are needed. In order to achieve
this, we focus on the problem of discovering spatiotemporal mobility profiles
from cell phone-based location logs. Approach: In order to capture the
mobility behaviors of cell phone users at a level of suitable abstraction, we
introduce formal definitions for the concepts of end locations (corresponds
to locations where users spend significant amount of time), mobility path
(denoting a user’s travel from one end-location to another), mobility pattern
(denoting a popular paths that users traveled frequently), and mobility
profile (providing a synopsis of a user’s mobility behavior by integrating
the frequent mobility patterns with time contextual data (2D location
component of profiles), and end locations with time distribution data (1D
component of the profiles)). These four different mobility concepts
correspond to abstraction of mobility information in different levels listed
from lower to the upper ones. Here, we design and implement a complete
framework, the Mobility Profiler, for discovering mobility profiles from raw
cell tower connection data by producing all of the data type abstractions
intermediate levels described above. Contributions:
Our
analysis of the cell phone users’ mobility behaviors yields important lessons
for networking researchers interested in testing large scale ad-hoc routing
protocols. As also identified in a recent studies, we found that users spend
approximately 85%of their time in 3 to 5 favorite locations, e.g., home,
work, shopping. However, our analysis has exposed more interesting phenomena
for the distribution of the remaining 15% of the users’ time. We identify a
significant long tail in a user’s location-time distribution: Approximately a
total of 15% of cell phone user’s time is spent in locations that each
appears with less than 1% of total time. Our another finding is that
significant amount of human mobility (85%) exhibits spatial and temporal
regularity where users move between their top-k locations. The regularity
property of these mobility profiles with time contextual information (days of
week, and hours of day domain) enables to develop different smartphone
applications such as early warning systems and route prediction applications
etc. Publications: ·
Murat
Ali Bayir, Murat Demirbas, Nathan Eagle: Discovering spatiotemporal mobility
profiles of cellphone users. WOWMOM 2009: 1-9 ·
Murat
Ali Bayir, Murat Demirbas, Nathan Eagle. "Mobility Profiler: A Framework
for Discovering Mobility Profiles of Cell Phone Users", (To appear at
Pervasive and Mobile Computing Journal, Elsevier). |
|
Figure 3: TRACK ME
Framework |
A Web Based Personalized Mobility Service for Smartphone
Applications Problem: Nowadays, most of the
basic web services use instant location information for providing suitable
content to the smartphone users. However, more intelligent smartphone
applications such as context-based search and advertising, early warning
systems, city-wide sensing applications may require additional information
about smartphone users such as their mobility profiles. In order to meet more
personalized demand of these applications more personalized web services
needed. Approach: In order to support
different smartphone applications with personalized mobility information of
cell phone users, we propose TRACK ME: A new web based framework for
smartphone applications with personalized lightweight mobility service as
well as location tracking and mobility profile construction. We showed that
our personalized mobility service support different smartphone applications
and it is lightweight enough to provide fast access to the mobility profiles
of smartphone users. Apart from personalized mobility service, our framework
also provides solution for location tracking and mobility profile
construction problem by employing Map/Reduce architecture. The proposed
framework separate heavy processing of mobility profile construction from
mobility profile access (which is extremely fast operation). Contributions: We showed that our
personalized mobility services support multiple applications such as location
prediction and air pollution exposure risk estimation. We also propose an
online solution to location prediction applications where it is possible to
predict future locations of smartphone user instantly by using query interface
that is provided by our mobility service. We illustrate that this application
is easily used for solving early warning problems mentioned above. For the
air pollution exposure risk estimation, we have showed that it is possible to
obtain more accurate risk estimation by using our mobility service than
residential based approach. Publications: ·
Murat
Ali Bayir, Murat Demirbas, Ahmet Cosar, "A Web Based Personalized
Mobility Service for Smartphone Applications" (To appear at The Computer
Journal, Oxford University Press). ·
Murat
Ali Bayir, Murat Demirbas, Ahmet Cosar: Track me! a web based location
tracking and analysis system for smart phone users. ISCIS 2009: 117-122 ·
Murat
Demirbas, Carole Rudra, Atri Rudra, Murat Ali Bayir: iMAP: Indirect
Measurement of Air Pollution with Cellphones. PerCom Workshops 2009: 1-6 |
|
|
PRO: A Profile-based Routing Protocol for Pocket Switched
Networks Problem: Delay
Tolerant Networks (DTNs), which are also known as intermittently, connected
networks, or opportunistic, store and-forward networks investigate routing
techniques in the environment where the connectivity is not exist all the
time. Recently Pocket Switched Networks (PSNs) have been formulated as a
subfield of DTNs where each node represents a person with a communication
device. Unlike the general DTNs, human factor plays very important role in
PSNs. The nature of human mobility and the structure of social networks
emerge as important factors in Pocket Switched Networks, while DTN routing
algorithms have been oblivious to them. Therefore, more context aware
(mobility profile and social network aware) routing protocols should be
developed for PSNs. Approach:
In this work, we are motivated by the observation that using smartphones it
is possible to maintain more detailed contextual information about the nodes
in the network, and hence design faster and more lightweight routing
protocols than the existing work on PSNs. Here, we propose a fast
(low-delivery-latency) and efficient (low-message-overhead) routing protocol
for PSNs, based on the regularity of human mobility profiles and of
intercontact events. Our protocol, namely PRO, (profile-based routing protocol),
is simple yet general enough to be easily instantiated to solve the
smartphone search applications. Contributions:
In a break from previous routing protocols, we showed that our protocol
treats node encounters as periodic patterns and exploit them to predict times
of future encounters. Our profile based estimation of intercontacts yields an
accurate ranking of the potential forwarding nodes as to their ability to
deliver the message earlier to the destination. We showed that PRO routing
protocol is completely decentralized and local to the nodes. PRO runs in an
adhoc manner and does not depend on any central infrastructure or third party
like Telephone Service Providers. Using the Reality Mining dataset, we
compare the performance of our protocol with most popular previous approaches
over both cell based mobility data (coarse granularity) and Bluetooth
connection data (fine granularity). Our experimental results showed that PRO
routing outperforms previous approaches in terms of end to end delay and communication
cost. Finally, we measure the performance of PRO on smartphone queries
described above and show that PRO achieves similar query performance with
Epidemic routing (in terms of delay and success) while using significantly
less communication cost. Publications: ·
Murat
Ali Bayir, Murat Demirbas, PRO: A Profile-based Routing Protocol for Pocket
Switched Networks (under submission). |
|
|
Crowd-Sourced
Sensing and Collaboration Using Twitter Problem:
Despite the availability of the sensor and smartphone devices to fulfill the
ubiquitous computing vision, the state-of-the-art has gap due to lack of
infrastructure to task/utilize these devices for collaboration and
coordination. We propose that Twitter can provide an “open” publish-subscribe
infrastructure for sensors and smartphones, and pave the way for ubiquitous
crowd-sourced sensing and collaboration applications. Approach:
We design and implement a crowd-sourced sensing and collaboration over
Twitter, and showcase our system in the context of two applications: a
crowd-sourced weather radar, and a participatory noise-mapping application.
Our system is composed of three components namely Askweet, Sensweet and
Twitter clients. Sensweet is a smartphone application that publishes
real-time readings from the integrated-sensors to Twitter. Askweet is a
program that listens to its Twitter account for questions and processes the
questions and aggregates the replies it receives to these questions from
Sensweet and the Twitter clients. Contributions:
We present an analysis of our real-world Twitter experiments to give insights
for the feasibility of our approach. We find that although we do not offer
the user any incentives to reply, our queries receive at least 15% reply
ratios. Surprisingly, 50% of the total replies arrive within the first 10
minutes of our query, and 80% of the replies arrives within the first 2
hours, enabling low latency operations for crowd-sourcing applications. Our
experiments also found that consistently the majority of replies come from
users that access Twitter from their mobile phones. Publications: ·
Murat Demirbas, Murat Ali Bayir, Cuneyt Gurcan
Akcora, Yavuz Selim Yilmaz, Hakan Ferhatosmanoglu, "Crowd-Sourced
Sensing and Collaboration Using Twitter". WOWMOM 2010. |
|
|
Identifying
Breakpoints in Public Opinion Problem:
In this work, we consider the problem of identifying breakpoints in public
opinion about certain topic. While traditional poll applications achieve this
by providing a snapshot of public opinion, they can neither track temporal
opinion changes nor capture opinions that are not asked in the questionnaire.
In order to overcome difficulties of these applications, we propose to use
Micro-blogs to capture changes in public opinion. Method:
In order to detect opinion changes, we integrate tweet based emotion vector
representation of time periods in vector space model and document based
representation of time periods. We show that the set space model can be used
to eliminate false positive opinion changes that vector space model finds on
a time domain. For representing events that caused changes in public opinion,
we propose a new scoring function to discover popular terms by considering
temporal dimension. Contributions:
By successfully combining these methods, we identified the time intervals
that include a change in public opinion over two case studies. From the
experimental results we found that using emotion corpus based method and set
space model methods together eliminates false positives and improves the
accuracy of breakpoint detection. We also create a customized news tracking
application that can notify users without flooding them with every new entry.
In this aspect, our application is superior to other services such as Google
Alert because we notify users only for significant events. Publications: Cuneyt Gurcan Akcora, Murat Ali
Bayir, Murat Demirbas, Hakan Ferhatosmanoglu, "Identifying BreakPoints
in Public Opinion", SOMA 2010, SIGKDD Workshop on Social Media Analytics. |
