Group Title: Department of Computer and Information Science and Engineering Technical Reports
Title: Gender-based grouping of mobile student societies
CITATION PDF VIEWER THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00095715/00001
 Material Information
Title: Gender-based grouping of mobile student societies
Series Title: Department of Computer and Information Science and Engineering Technical Reports
Physical Description: Book
Language: English
Creator: Kumar, Udayan
Yadev, Nikhil
Helmy, Ahmed
Publisher: Department of Computer and Information Science and Engineering, University of Florida
Place of Publication: Gainesville, Fla.
Copyright Date: 2008
 Record Information
Bibliographic ID: UF00095715
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.

Downloads

This item has the following downloads:

2008447 ( PDF )


Full Text











Gender-based Grouping of Mobile Student

Societies


Udayan Kumar, Nikhil Yadav, Ahmed Helmy
Dept. of Computer and Information Science and Engineering, University Of Florida
Gainesville, Florida, USA
Email: {ukumar, nyadav, helmy}@cise.ufl.edu


Abstract The next frontier for sensor networks is
sensing the human society. Several mobile societies are
emerging, especially with wide deployment of wireless
LANs (WLANs) on campuses. With the rapid increase
in WLAN deployment come various research
challenges. WLAN traces pertaining to network usage
contain very useful information and can provide us with
a lot of insight about mobile user behavior and network
usage. Such insight may be used to design better future
networks, and analyze the effect of social attributes on
usage of mobile technology. The most extensive libraries
of wireless traces are collected from university
campuses. The traces are anonymized and do not
provide affiliation or preference information explicitly.
Hence, it becomes a challenge to perform social or
group-based analysis with existing traces. In this paper
we present a novel technique to group WLAN users
based on gender. By mapping the traces into buildings
(including sororities and fraternities), we then extract
affiliation (and hence gender) information based on
statistical network usage. Once such grouping is
attained, we examine the commonalities and differences
of usage patterns and preferences between male and
female groups. Parameters analyzed in our study
include on-line session durations, vendor preference,
and user distribution on campus and across study
majors. Our results clearly indicate the effect of gender
on on-line behavior and vendor preference. We find
such effect to be statistically significant and consistent
across various semesters. Our findings provide great
promise in utilizing WLAN traces for future mobile
applications, yet raise privacy concerns that we plan to
investigate in future work. Also, our method provides a
framework for analyzing group behavior in mobile
networks in further studies.

Keywords- Social grouping; group behavior; privacy;
WLAN

I. INTRODUCTION

Most existing studies on sensor networks focus
on sensing the physical world and phenomena.
Although very useful, there have been very few (if
any) studies on sensing the human society, which
presents a new set of intricate and intriguing research
questions. In this study we present one attempt to use


WLAN traces to better understand certain behaviors,
based on groupings, in wireless mobile societies.
There has been a rapid increase in WLAN
deployment across university campuses. User activity
on these networks is growing dramatically. As
student societies become more mobile, with the
ubiquity of wireless coverage and availability of new
portable devices, there are great research
opportunities to mine and study mobile student
societies to understand their behavior, preferences,
grouping, among other characteristics. Such
understanding opens the door for the efficient design
of future networks, and has been facilitated by the
availability of recent libraries of WLAN traces [8][7].
The traces provide anonymized individual traces, but
lack any information about the social context,
attributes, affiliation or gender, and hence hide
potentially very interesting characteristics of group
behavior in mobile societies.
While previous works studied WLAN
deployment issues [1], issues of mobility [2] and user
association patterns [3][4], we aim to address issues
of user classification based on social grouping. In this
paper we set out to analyze WLAN usage patterns
based on gender, majors and other interest groups.
This allows us to examine trends among different
social groups. Having an insight into user behavior
from social standpoint can give us a lot of input in
designing network protocols, such as delay tolerant
networks (DTNs). Mobility analysis [2] and user
predictions in mobile networks already gain a lot by
incorporating user's social behavior. Context aware
services [11] of mobile networks of the future would
need to understand the context from user's
perspective, so they may have to understand the
social behavior of the user.
We propose to use WLAN traces, which are
generally considered for studying network
characteristics, to mine social behavior of the users.
We present a general methodology with an example
case study of grouping by gender, and investigate
gender gaps in WLAN usage. The lack of such
empirical data poses an interesting challenge and















raises several research (and privacy) questions: How
can we meaningfully infer gender information from
such anonymous traces? Will such information
influence user behavior and preference in a
significant and consistent manner? In this paper, we
introduce a novel technique to mine WLAN usage
patterns based on gender, majors and other interest
groups. Some of the central ideas in our paper include
the user of building map information for the WLAN
traces, knowledge of locations of departments,
fraternities and sororities, and the use of sound
statistical methods to classify users in majors, males
and females and reason about the effect of such
classification on network activity and preference. The
method we provide can be used further to analyze
behavior based on various other groupings.
Gender based studies have been conducted in the
past to study issues such as the difference in
technology adoption for the Internet [5]. This paper is
the first, to our knowledge to analyze WLAN
adoption patterns across these groups. Among the
parameters we have considered for evaluating the
gender gaps, we found enough statistical evidence to
conclude that (for the traces in our study) usage
patterns of males and females is different, and that
gender does affect user activity and vendor
preference. Our success also indicates that the
problem of mobile user privacy should be re-visited.
The rest of the paper is outlined as follows. Section II
discusses the main challenges in our study. Section III
provides our approach for data analysis and our
method. Section IV provides details about the traces
used in our study, and Section V presents out data
filtering technique for gender classification. Section
VI provides the gender-based analysis and results.
Section VII. Discusses potential applications and
Section VIII concludes.

II. CHALLENGES

How can we begin to classify all the students into
groups like gender and study major using only the
publicly available information? The method we have
used involves processing and analysis of the WLAN
traces. Traces are logs of user association with
wireless Access Points (AP). Traces generally contain
only user's machine's MAC ID, associating time,
duration and associated AP [7][8]. Often, because of
user privacy issues, the MAC IDs are anonymized.
Having a meaningful classification with this partial
information is the main challenge we address. Ideally
we would want to classify all students into groups.
Taking a first step in this direction we present a
general technique which can be used to classify a
smaller section of WLAN users into groups. Doing it
for the all students still remains a challenge as we


shall see. Instead we focus on obtaining a sample
significant enough for a statistical analysis.


ISut Tables on which further SQL queries can be run)
Fig 1: Query based User grouping Technique


MA START I AP I DURATION I MNU C | BUILDING I XI Y
WI W 515101 172168245 31021 ag Ind pMd 2
N53T765 172 61 245 2113 95 iniel a- 4 2
I n 53 9451 17216824521005 115 Inlrd hoh 5 1


III. APPROACH an21d METHOD of DATA



ANALYSIS

Our technique works on raw WLAN SNMP and
31SYSLOG traces. The traces are accumulated for the
4*AP 37SM518 172-6245_2102 100 5 We s 4 4
5141088 n16S242J11 071 1 592 Appe pkt M M
a 6622 72 831104 43B -1 P O. .1 gI
Fig 2: Trace Database
1II. APPROACH and METHOD of DATA
ANALYSIS

Our technique works on raw WLAN SNMP and
SYSLOG traces. The traces are accumulated for the
period we are interested in studying and parsed into a
standard format as in Fig 1. We also use the location
information of the APs, in the form of buildings in
which they are located. This helps in knowing the
geographic locations of a user at a later stage.
Mobility of users can be tracked by looking at the
approximate geographic locations of the APs.
The processed data is fed into a database on
which SQL queries can be run easily (and
generically) to extract information of interest to us.
Fig. 2 illustrates the trace database layout which was
used in our experimentation. The fields include the
following: 1. MAC IDs of the wireless devices logged
onto the WLAN, 2. the starting session time in
seconds, 3. the AP which wireless device logs into,
4. Duration in seconds that the MAC stays logged
into the AP, 5. the manufacturer (which can be
inferred from MAC ID), and 6. the building the AP is
located at (approximately), which can be checked
based on access point location information which is
external data to the actual traces. Two-dimensional
co-ordinates can be inbuilt into the database based on
a campus grid map to allow mobility based queries to
be performed as well.














The trace database provides enough information
on which to run SQL queries. For instance, a simple
query returns the number of MACs logged into
building a or b with durations within a certain range.


Fig 3: Gender grouping in Fraternities and Sororities


We have used this same database framework to
analyze traces from USC[8], Dartmouth [7], UF and
UNC[10], the method is general and applicable to
many traces, campuses and societies. Completing
these analyses is part of our future work.
The grouping parameter we use for investigation
in this document is gender. To do this categorization,
we propose the following novel technique. Most
universities have sororities and fraternities as social
organizations. Sororities are female organizations
while fraternities represent male organizations. Given
the physical location of APs on campus, APs located
in sororities and fraternities are identified, and the
users associated with them are classified as female or
male. Fig. 3 shows how grouping is done in this
setting. The fact that visitors may frequent these
locations also needs to be taken into account. We deal
with visitors in the filtering section.


IV. CHOICE OF TRACE

In order to carry out a meaningful analysis of the
various groups of WLAN users, specifically targeting
the differences (if any) between genders, the traces
need to provide the following information:
i. comprehensive Syslog (or SNMP) logs for various
buildings, dorms and sororities/fraterities for the
general population of students (without population
bias) and for extended periods of time (30 days or
more), ii. mapping between the APs (or point of
collection) to specific building's designation.
iii. differentiation between individual users and
ability to track the same users along the whole
duration of trace (without necessarily knowing their
identity). We have investigated WLAN traces
collected at USC[8], UNC and Dartmouth[7].
Dartmouth traces do not provide AP-to-building
mapping, which makes it difficult to do this kind of
study. UNC traces on the other hand have limited


number of APs in sororities and fraternities. We
chose the USC traces for our study as 12 fraternities
and 7 sororities are included in WLAN traces and the
AP-to-building mapping is also available. We have
chosen 3 months for study Feb 2006, Oct 2006 and
Feb 2007. The reason for having traces from multiple
periods is to look at consistency in the results and
also at the trends. Traces have been taken from
different semesters, so if there is something as a
semester effect in the results it should also be visible.

V. FILTERING of VISITORS

As fraternities and sororities have male and
female visitors, without further refinements and
filtering our classification would not be very accurate.
But even if we validate the presence of visitors, how
can we remove them from our classification? First,
visitors are infrequent users of the mobile network in
the visited locations. Second, we expect a significant


Fraternity feb2006




i


Distlnct Users
(a) Feb2006











(b) Oct 2006



difference between residents and visitors in terms








Fig 4: Session count for Sorority and Fraternity users














network activity (in number and duration of on-line sizes. An int
sessions). Hence, we define a visitor as a user with that the sha
less number of sessions and smaller duration of stable after t
sessions than the average user in that location, the suitabili
Therefore, our technique rates users based on two duration in s
metrics, the number of sessions and session duration.
Fig 4 represents session counts per MAC ID in VI. ANAL'
decreasing order (for more detailed analysis see[6]).
Fig 4 graphs are produced using the average session
duration (in sororities and fraternities, respectively) as Once w
the threshold for session duration. We observe users, we c,
interesting, distinguished characteristic in Fig 4 that campus tract
indicates the presence of a sharp bend (knee) as the pattern over
number of sessions per MAC ID decreases. investigate
Intuitively, this means that MAC IDs below the knee questions:
have an order of magnitude less number of sessions a) WL
(accounting for the difference between a regular user Wh
and a visitor). All users below the knee were diff
classified as visitors and removed from the study. b) Ave
While users above the knee in sororities (fraternities) ave
were classified as females (males). Changing the diff
session duration had no effect on the shapes of the (bu:
curves [6]. c) Mai



160
7 Male
140- m Male
0 120 Male
le_ T Femn
100 Fem
S80 lFemi
60
.i40
D O

20



0N Area


resting observation to note from that is
pe of the curves (with knee) becomes
he 4th day of the trace [6]. This indicates
ty of our analysis to traces of shorter
similar environments.

YSIS OF MALE AND FEMALE WLAN
USAGE

e have the MAC ID's of male and female
an analyze their behavior in the whole
Sand identify and characterize their usage
the whole campus. In this analysis we
the following usage and behavioral

AN Usage and Gender Distribution:
at are the trends in WLAN usage across
erent (buildings) areas on campus?
rage online time: Are there trends in the
rage online times of users and can
erences be spotted based on gender and
ilding) areas within the campus?
nufacturer preferences: Which device


feb 2006
oct 2006
feb 2007
ale feb 2006
aleoct_2006
ale feb 2007


Q00iipe~ &~Oa\ 50~
so S
S

Fig 5: Unique user count across different areas


The general steps for visitor filtering are:
i. Extract the number of sessions per MAC ID for
each fraternity or sorority AP. ii. Vary the minimum
session duration (as a threshold for regular users) and
observe its effect on the number of sessions and
distinct users. iii. Obtain a suitable threshold for the
session duration and session count to classify users
above these limits as being either males or females.
We also performed a time evolution study for
number of sessions per MAC ID varying the
minimum average session duration. In such study we
perform the filtering for decreasing sample/trace


vendors do different genders prefer?

A. WLAN usage by area

We track MAC ID's of the previously
identified/classified male and female students in
different areas of the campus. If the session duration
of the user is above a threshold (corresponding to the
average user in that location), we consider that user a
regular user of this area. Fig 5 shows the usage
distribution per area type based on our definition of
'regular user'. Economics buildings show a higher
















Male feb 2006
Male oct 2006
Male feb_2007
= Female feb 2006
[II Femaleoct_2006
=Q Female feb 2007


GO Area
Fig 6: areas Average Session duration by Area


population of male users, social science buildings
have a higher count of female users. It is interesting
to see that there are more female WLAN users than
males in Engineering buildings for the sample Feb
2006 however Males users take the lead by Feb 2007.
We see that absolute number of students classified as
male and females increase in Oct 2006 and then drop
down in Feb 2007. This is perhaps due to the fact that
more students join in the Fall semester than in any
other period of the year and also that many students
graduate after the Fall semester. It also indicates that
several courses offered in the Fall require the use of
laptops (and the wireless network) for the course
work.

B. Average session duration

We now study the average session duration for
male and female users across campus. From Fig. 6 we
observe that males spend more time online than
females in most of the areas. Females show dominant
usage in the Social Science, Economics and Medicine
areas across campus. We can deduce from this that on


average, male users tend to stay as WLAN users at
certain places for longer times than females. Another
observation of interest is that average duration per
session decreases from Feb 2006 to Feb 2007 in
almost all the cases (Engineering, residence, social,
sports, music). This points to the possibility that
students are becoming more mobile, and thus having
shorter sessions in the same location.

C. Manufacturer Preferences

The preference of manufacturer (based on the
type of wireless card traced) is shown In Fig.7. It is
interesting to note that Apple computers are more
popular amongstfemales than males. Intel devices are
more popular amongst males. For this study, only
major vendors were considered. For example using
the Feb 2006 we find that: In case of males there are
~25% using Apple and ~32% using Intel, so there are
28% more male users using Intel with respect to
Apple users. In the case of Females: there are ~30 %
using Apple and ~27% using Intel, so 12% more
females users using Apple with respect to Intel users.


I ,,, i =I., I. ,.
I ,,' ,-_ i, .'I, .-, ')
II. ,,,, I,- l l. -I. H)


-, I-,


__2 2-2-r1 LfWL I I' I''


Apple


Intel


Gemtek Enterasys
Manufacturers


Linksys


ASKEY corp.


Fig 7 Device distribution by manufacturer


3000-

2 2500-
0)
S 2000-

1500-

0D 1000-
50
0 500-
cc


o


St


I














To test whether gender provides a bias towards
specific vendors, we use the statistical significance
test, Chi-Square. The Chi-Square test shows with
90% confidence that there is a bias between gender
and vendor/brand.
Another interesting observation we make from
Fig 7 is the consistent trend of increasing percentage
of Apple computers usage in both the genders. We
also see that vendors like Enterasys, Linksys and
Askey Corp. have a decreasing trend in terms of
percentage of users. One of the reasons is that these
manufacturers mostly make external Wi-Fi devices
for old laptops (with no built-in Wi-Fi NICs).
Currently almost all new laptops come with a built-in
Wi-Fi, so the users of external devices are
decreasing.

VII. APPLICATIONS

In general our method for user classification and
grouping can be used to group users. It may be used
to profile users, grouping them based on their on-line
activity. The current gender based analysis can be
used to find out the extent of WLAN adoption
amongst genders, which is of great interest to social
scientists studying socio-economic and socio-cultural
differences between genders. Further investigation of
group behavior can be used to predict user
movements. Trends in mobile social networking for
these groups can be used to provide specific services.
Protocols can be made context aware, if they have
access to such user classification methods.
Announcements and advertisements on campus can
be directed based on the general psyche of males and
females. The areas these users frequent more could
serve as good places to advertise related interests,
services or products. This is part of a paradigm our
group is designing called Profile-Cast [9].

VIII. CONCLUSIONS AND FUTURE WORK

In this study we propose a novel technique to
classify WLAN mobile users into groups by
analyzing anonymized WLAN traces from a major
university campus. We utilize mapping information
of buildings and departments to obtain a meaningful,
statistically sound classification. We focus our
analysis on gender-based groups. Results from this
research are based on a sample of the user population,
since gender may be identified based on sorority and
fraternity wireless access point associations. We find
that there is a distinct difference in WLAN usage
patterns for different genders even with similar
population sizes. Females seem to dominate in
WLAN usage in areas of Social Science and
Economics and prefer Apple over Intel. Males have


longer session durations than females in most cases.
We see that these trends and characteristics are
consistent over periods of time and across different
semesters.
Our methodology of gender classification and the
use of SQL queries on the WLAN traces are generic,
and can be applied to classify users into groups like
study major and various interests. In our future work,
we plan to study similar characteristics across
different university campuses, including a detailed
time-based analysis of gender mobility based on
different time periods of the day.
We interestingly note that we were able to
classify users into male and female and were even
successful in obtaining their preference of vendor,
based on analysis of anonymized traces. Our study
was based on group (not individual) behavior. Yet
there are several privacy issues raised implicitly in
our work. Can private information of individuals be
identified by analyzing anonymized traces? What
kind of anonymization algorithms should be used for
mobile networks traces? And how can such
algorithms provide a notion of k-anonymity for the
mobile society while retaining useful information for
researchers? These are questions that bear further
research and we plan to address them in our future
work.
We hope for this study to open the door for other
mobile social networking studies and profile-based
service designs based on sensing the human societies.



REFERENCES

[1] T. Henderson, D. Kotz and I. Abyzov, "The
Changing Usage of a Mature Campus-wide
Wireless Network," in Proceedings of ACM
MobiCom 2004, September 2004.
[2] W. Hsu, T. Spyropoulos, K. Psounis, and A.
Helmy, Modeling Time-variant User
Mobility in Wireless Mobile Networks ," in
Proceedings of IEEE INFOCOM, May 2007
[3] W. Hsu and A. Helmy, "On Modeling User
Associations in Wireless LAN Traces on
University Campuses," The Second
International Workshop on Wireless
Network Measurement (WiNMee 2006),
Boston MA, Apr. 2006.
[4] G. Chen, H. Huang, and M. Kim, "Mining
Frequent and Periodic Association Patterns,"
Dartmouth College Computer Science
Technical Report TR2005-550, July 2005.
[5] Ruby Roy Dholakia et al. "Gender and
Internet Usage, "The Internet
Encyclopedia", Wiley, 2003.














[6] http://nile.cise.ufl.edu/socnet
[7] CRAWDAD is the Community Resource for
Archiving Wireless Data At Dartmouth
http://crawdad.cs.dartmouth.edu/data.php
[8] MobiLib: Community-wide Library of
Mobility and Wireless Networks
Measurements (Investigating User Behavior
in Wireless Environments).
http://nile.cise.ufl.edu/MobiLib.
[9] W. Hsu, D. Dutta and A. Helmy, "Profile-
Cast: Behavior-Aware Mobile Networking,"
ACM MOBICOM poster and student
research competition, Montreal, Canada,
Sep. 2007
[10]UNC/FORTH repository of traces and
models for wireless networks, Syslog dataset
#2, http://netserver.ics.forth.gr/datatraces/
[11]Wei-jen Hsu, Debojyoti Dutta, and Ahmed
Helmy, "Profile-Cast: Behavior-Aware
Mobile Networking," to appear in IEEE
WCNC, Las Vegas, NV, Mar. 2008.




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs