A Comprehensive Technical Review on Social Network based on Call Detail Record

munchsistersΤεχνίτη Νοημοσύνη και Ρομποτική

17 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

83 εμφανίσεις

A Comprehensive Technical Review on S
ocial Network based on Call Detail

Record

Preetish Ranjan, Abhishek Vaish

Abstract:

Social Network Analysis is emerging as a sought after discipline for researcher due to the fact that it has big data
and gaining
insight on that is a challenging task. The scope of the social network does not limit to the social media
but also mobile communication which is connecting the entire globe

that has similar quantum of data for analysis
.
The interesting issue here is
that t
he later and its

analysis is not adequately addressed

as a subject matter of study

and therefore a comprehensive re
view is required
for the researcher to understand its dynamics and to conduct
research in this direction. This paper presents a very exhausti
ve survey and focuses
to explore the entire
gamut of
the mobile pho
ne generated data and its usage
such as exploiting

user behavior, social interaction through calling
pattern, marketing strategy, transportation mode used by travelers etc. One more
dimension

gaining the maximum
attention

in the sphere is the usage of mobile network for planning and coordination of criminal activities which
ultimately converge to the phenomenon of social technical attack.


Keywords: Social Network, mobile communicati
on, mobile phone, user behavior, calling pattern, social technical
attack


Introduction
:

Un
doubtedly
,

the
power of mobile communication has over ruled all the communication technologies

i.e. i
n the past
,
irrespective of the fact of the geo
-
position,
socio
-
economic developm
ent,
technological advancement o
f the
individual country,
the adoption of the GSM/CDMA based mobile communication is phenom
enal and penetration of
the same
can be seen in fig

2
. Some of the most prominent reason for this growth is th
e inexpensive

cost of
technology, the benefit
of the low cost
of
service and the ease of getting connected to the whole world.
Mobile
phone has become part and parcel of our life and it has become easier to communicate with each other and share
each and ev
ery moment of our life.

It has become an integral part for every individual and now people prefer
multiple mobile sets or dual SIM set that lead to a revolutionary transformatio
n in every aspect of human life.

Since
the mobile phones have become an importa
nt tool for modern human daily life, hence, it is being used as
notepad,
calendar, calculator, alarm clock, reminder

etc.

Some are using to maintain their social relationships with near and
dear ones, parents are using it to be in touch with their wards, students are accessing internet through it to enhan
ce
their knowledge, air
and flight tickets may be booked and money may
be transacted through it. But on the other
hand,

there are some negative aspiration
also such as disturbing others with the false phone calls and corroborating
crimes such as kidnappings
, sending bulk SMSs and MMSs to spread wrong messages to hurt the feel
ings of mass

etc.

The paper is focused to bring a plethora of
A
pplied side of

Mobile Communication Usage and the CDR
Analysis.








Fig.1

An exhaustive survey which followed 94 subjects using mobile phones preinstalled with some software that
recorded and sent the surveyor the information about call logs, Bluetooth devices within 5 meters, cell tower_id,
application usage and status. Reactio
ns of different user

s

group

such as to

answer or not to answer, inf
ormation of
individual’s desire and
call handling decision have been observed through many surveys with

the support

of
different methodologies.

It has been observed that user’s reactions
may be broadly classified in three categories.










Fig. 2

Every respondent believes that the primary function of mobile phone is to make a call whereas texting
is the
second
most important function.
General users about
67% don’t bother about the netw
ork they are making
call
to and less
than half (45%) know which network they call most but 27% share the same network. 70% of users do not know the
cost to call another mobile number and 75.3% connects to internet through their mobile phone technology
[1]
.

Youth mostly communicate within their age group to discuss their problem related to their place, time and purpose.
They prefer mobile phone to remain in touch with friends and families as surveyed that 2/3 of Americans aged
between 16 to 29 years choose mo
bile phone a
head of land lines but
31% of those aged over 50 years prefer
landlines to talk. They are mostly concerned with very low priced prepaid packages offered by telecom operators
and instant texting starts after the age 10 or 11 especially among gir
ls. One study in UK few years back came up
with
a
very surprising hypothesis that the percentage of smokers among teens has dropped claiming mobile phones
are proving substitute for this habit

[1]
.They use latest mobile technol
ogies as a mean

to connect to

i
nternet to access
information. Fig.3, Fig.4 and Fig.5 depicts the percentage of users from different section of society using mobile
phone device as per their perspective.















Fig.3

Fig.3
deduce that atleast 1 to 10 calls are dialed or received daily especially by homemakers but 11 to 20 calls are
dialed or received by businessman.








Fig.4

Fig.4 summaries that guardian
s
/relative
s

are most active in making mobile communication and
everybody want to
remain connected to those who are somehow related to their profession.





Students

Homemaker

Businessman

Serviceman

Total
























Fig.5

Fig.5 depicts that students prefer texting and cal
ling over night most but most of the user except student do
not
prefer to call over night. After student it is a serviceman who usually like messaging and calling over night.


Extraction of
buried
information

f
rom

CDR (
C
all
D
etail
R
ecord
)

Some people are trying to carve a way from CDR to identify the cases of fraudulent usage, abnormal behavior
patterns, research study
and the problems in the billing process
etc

[42]
. But, others are interested to find the outliers
on the basis of character
istics of the shape of the time series.

In the paper titled as “Analyzing and Geo
-
visualizing
Individual Human Mobility Pattern using CDR” introduced several preprocessing and spatiotemporal analysis for
human mobility pattern mining using 2D or 3D graphic

representations of individual activities

[11]
.
Recent research
after the analysis of CDR revealed that the most frequent mode of the activi
ty is related to their

occupation

[14]

and

some paper presents an outlier detection algorithm based on patterns
derived from the skeleton points of time series

using pattern descriptor to capture the main characteristics of the shape of the time series [38]
. Modeling a call detail
record while collecting and storing data by different mobile operators affects the bil
ling process
as well as to
leverag
e

missing and corrupted values during fraud analysis on the basis of probability, significance and
Naïve
Bayes Classifier [45]
. A very unique experiment has been performed on CDR to identify the transportation mode
shared
from given origin to given destination on the basis of mobile’s phone speed wh
ich may be usually estimated
through
the variance of GSM signal strength or the switch rate of cells. If the estimated speed is within certain
range, then it is believed that th
e mobile phone user is in a specific transportation mode and the pe
rcentage of
travelers using different

transportation facility may also be determined

[48]
.

In the paper titled as “Anatomy of Data
Integration” initiates the framework to improve the data q
uality
for accessing the informational value and facilitating
integration of data from diverse source to encourage

better
business practices

[18]
. In another paper, a new fault
prediction model
based on either incomplete or vague

input

is being focused

to
deal with
fault prediction problems
in the presence of both qualitative and quantitative
data

[19]. Mobile call
graph is being investigated to draw the
distribution

curve which
may follow

fundamental
power law and lognormal distribution as most of social network
graph follows
this
. CDRs are quite frequently used for unsupervi
sed machine learning
to identify the clusters based
on user behavior in telecommunication network using the social features inste
ad individual features [34].
Apple’s
Shark is a statistical call path profiler
which implements

the
different
profile of call path

on the basis of stack sample
of frequency counts of call graph edge

and it

does not
really
deal with

the content of call [41]
.
There is a case study
from Telecom Enterprise for examining the performance and reliability improvement of CDR processing system
through job scheduling with dynamic priority assignment [47].


Use of call logs

in long term business decision

Not only telec
om sector but other b
usiness houses are using call logs to
know about their customers and accordingly
decide their future investment
s. It may be classified in four categories as follows









Fig.6

Marketing
professionals are analyzing call detail record to evaluate the increment in mobile

phone

user
in coming
period.

Machine learning technique is being used
to predict the churn of customer switching from one service
provider to other

based on calling links as input to a neural network model [13]
.

Various data mining techniques are
being implemented to assign a ‘propensity
-
to
-
churn’ score periodically to each subscriber of a mobile operator.

Technique based on the multi
-
classifier clas
s combiner approach has been proposed to identify potential churners for
a specific period

which addressed the challenge to identify the highly skewed class distribution between churners
and non
-
churners [16]
. Telemarketer
use to
i
mport set
s of CDR

into da
tabase for data analysis

to identify their
customer base for expanding their business [32].

A model is being proposed based on the

relevancy of local search
and
probability of calls made to particular categories of business as a function of distance by com
p
aring click log
and mobile log [33].

Few t
elecom operators decide their pricing strategy by analyzing customer’s call
ing

behavior
patterns obtained from CDRs

[28]
.

In the paper titled as “Calling
Communities Analysis and Identification U
sing
M
achine
L
earning
T
echnique” focuses on how to
identify the cluster of customers

using m
achine learning technique
based on their

calling pattern.

These clusters have been classified for more information with optimum classification
accuracy and computational performa
nce [43].


Use of CDR in Social Network Analysis (SNA):

Telephonic interactions form
stronger

network than that of online social interactions, therefore it requires

more
attention. R
esearch are trying to analyze
this telephonic social network

and concentrating on following areas








Fig.7

The effectiveness of different machine learning techniques such as Support Vector Machine and Fuzzy
-
genetic
classifier is being studied to generate a model for automatic identification of different callin
g groups of customers
[8]. Mahalanobis, Euclidean and Hellinger distance is being implemented among nodes in the form of mobile
number to find the closeness and social tie strength among users and graph theory, game theory and the concept of
various data m
ining algorithm such as association rule, Naïve Bayes algorithms is also being used to analyze as well
as predict the social network structure. The paper titled “Predicting Social Ties in Mobile Phone Network”
investigates the evolution of person
-
to
-
person

social relationship, quantify social tie strength, level of reciprocity
between user and their communication partners and calculated Hellinger distance to predict the social
-
tie among
mobile phone users [3]. According to the results of some experiments to

measure the dyadic relationship and
mutuality, it has been observed that the higher reciprocity index value reflects a closer relationship between
members and this is useful for detecting unwanted calls and spam marketing etc [4]. The case study on the Ca
talano
Phone Call represented a large class of visual analysis operations on graph data by exploring a geographical location
and movement of mobile phone user [31]. Time varying relationship in statistical relational model based on dynamic
graphs is being
exploited with a weighted static graph and then it is being concatenated with link weights in a
relational Bayes classifier [36]. There has been a very innovative approach for efficient proximity search in
multivariate data for deriving a new class of hier
archical metric data structure by applying multi
-
polar mappings to
metric data and for hierarchical decomposition in multi
-
dimensional space to spatial data structure [37].


Evaluation of subscriber by telecom

service provider

Calculation of propensity to churn
by customers

Pricing strategy by telecom
operators

Marketing survey

Use of CDR in long
term business
decision

Exploiting social
interaction

Quantifying
reciprocity in social
network

Prediction of social
tie strength

Calculating
willingness level of
the receiver

Fraud detection
:

There
is different kind

of frauds prevailing in our society

but t
here is different class of frauds which are committed
through mobile phones and it may be classified as follows







Fig.8

Mobile phones are
being used by the fraudster as a mean of gaining free access to a cellular t
elecommunication
network
by
reprogramming the embedded software

to transmit the electron
ic serial and telephone number
of
authentic user.

The paper titled as “Adaptive Fraud Detection” tries to detect the cellular cloning fraud through user
profiling method using data
mining techniqu
es, r
ule learning program and setting the
indicators to indicate
anomalies in database

[30]
.

Due to easy availability of mobile phone and SIM card
, it is readily used to plan and corroborate crime such as
kidnapping as it help to transfer complete informat
ion of each and every move at the same time to any distant place.

The human mobility is always being ver
y dynamic and deriving pattern from it is too complex

but one

paper have
tried to analyze these trajectory data to unveil the complexity of human mobility through querying and mining to
resolve the

raw GPS track in to mobility knowledge discovery

by describing M
-
Atlas [39]
.

There are some paging
scheme to
determine
how and where to
search for a mobile user

in pool of data

on the basis latest location update
information from that user and developed a family of profile based

on
paging technique which are updated quite
frequently.

They concluded that highly mobile user
s tend to make and receive
more calls than static users and
entropy of location is being calculated simultaneously [40].
There
is

some call path profiler based on the stack
sampling which uses novel sample driven approach for collecting frequency counts of

call graph edges.

While
exploring a calling path patterns

using data structure of path log
,

a algorithm is being designed to mine calling path
pattern

using the constraint of limited number of neighboring cells

and
it has been diagnosed that a vertex in
the
GSM network graph has at most 36 in
-
out paths and six out
-
edges

[21]
.


Fraud such as impersonating using other’s identity to make a call is very frequent and few researchers are working
to tackle this kind of problems

using back propagation neural net
work

to perform telecommunication interpolation
based on local network

[10]
.

Band width is always being a scar resource

in telecom sector,

hence
,

the problem
of
resource reservation and call admission control in wireless mobile network has been addressed
i
n many papers. By
analyzing the previous, local and global movement profiles for mobile users
, the future movement path may be
predicted.

Anomalous users are being detected based on fuzzy attribute values derived from the communication
patterns using CDRs

to identify differ
ent communities of users and
assist human analyst in validating the result [25].


Use of CDR in investigation p
rocess

Investigating agencies and police departments are frequently asking the CDRs of particular user during period of
crime
and analyzing it to identify the evidence which may be proved in court of law
. Evidence are based on the
following information
:









Fig.9

Investigators are trying to develop a trace driven simulator to analyze the behavior of mobile systems in a
social
environment by capturing the specific characteristics of friend and stranger encounters

[2]
. There is a huge
significance and impact of missing values in the CDR which may be due to a software or manual fault but some
papers are targeting to leverag
e missing values using Naïve Bayes posterior to rule based classifier to analyze the
probability of corrected and missing values in CDR

[45]. The paper titled as “The random subspace binary logit
model for bankruptcy prediction” proposes a random subspace
approach model to generate the sets of variables to
represent the problem of bankruptcy. It has been tried to localize mobile phone based on knowledge of the network
layout and incorporate additional information such as round trip time, signal to noise rat
io etc

[20]
. Some papers
emphasizes on the data generated by telecom industry may be useful for the law enforcing agencies to find out the
network of some criminals

through social network analysis and by calculating the prominence of an actor in the
networ
k [35]
.

There has been a regression based approach for mining user movement pattern from random sample
data from CDR and the result of this experiment involving both synthetic and real data sets is very close to frequent
movement behavior of a mobile user

[12]. The technique based on Bayesian approach is being implemented along
with round trip time and signal to noise ratio to locate mobile phone users in the cellular network layout and result
have shown that this technique can reduce the localization erro
r by 2% compared to blind approach [29]. Criminals
may be tracked by executing proposed guessing program of PathID and VPI/VCI (virtual path identifier/virtual
channel identifier) mapping [44].


Methodologies used to analyze call logs

After going through number of paper, it may be inference that
either theory of probability or concepts of statistics or
data mining algorithms are being implement or cumulative result of these three have been used to predict the future
trend. It may be rep
resented as
follows:



Fig.10

Some paper propose
d

to predict user mobility statistically based on the mobility history of users and analyzing the
sequence of events of
new call admission, hand offs and call termination
. These variables

are bei
ng modeled by the
stationary m
th

order Markov sources

which derive the probabilistic prediction of next event based on the mobility
history of user [17]
.

Call a
dmission and resource allocation by the user will be based on the previous movement,
local and g
lobal mobility profiles of the user

which may be further
utilized effectively in the pre
diction of future
path of a mobile user [
15
].

Now, call logs are being used
in
security and management in IP telephony

to support
researchers and practi
ti
oners

t
o
handle different non
-
standard logging format across operators and underst
and the
com
mon source

of call log errors to streamline their investments in monitoring system by collecting valuable
information

[27]
.

ARIMA model is being implemented over CDR to for
ecast the increase in the number of mobile
phone user in coming years

[1]
.



Conclusion:

After a very exhaustive survey it may be concluded that
in this

era of technological advancement, the usage of
mobile phone device
s

depends on the perception of particular person or group
. The magnitude of its usage
eventually depends on the domain and the challenges associated with the domain.

It has been observed that
mobile
phone is indispensable part of our life
-
style irrespective

of age, gender, profession,

salary, expenditure etc.
Instant
messaging begins at the age of 10 or 11, most of the homemaker and student use pre
-
paid connection but post
-
paid
connections are preferred by service holder
s
.

I
t has been inferred that
most cal
l
s

are made to person by profession of respondents and this give rise to the
development of some kind of social network among mobile phone users.

This social network is being analyzed
through CDR to extract hidden information
and apart from this
,

CDRs help in
research study, mobility pattern
mini
ng, identification of fraud,
long term business decision and investigation process.
Probabilistic, statistical and
various data mining algorithms are being used to extract useful information from CDR but t
here still exist certain
challenges

of false positive or true negative results

which give the scope of further research.






References:

1.

http://www.itu.int/osg/spu/ni/futuremobile/Youth.pdf.

2.

Andrew G. Miklas, Kiran K. Gollu, Kelvin K.W. Chan, Stefan
Saroiu, Krishna P. Gummadi and Eyal de
Lara, “Exploiting Social Interactions in Mobile Systems”, Springer
-
Verlag Berlin Heidelberg 2007, pages
409
-
428.

3.

Huiqi Zhang, Ram Dantu, “Predicting Social Ties in Mobile Phone Network”, IEEE, 2010, Pages 25
-
30.

4.

Huiqi

Zhang, Ram Dantu, Joao Cangussu, “Quantifying Reciprocity in Social Networks”, IEEE,
International Conference on Computational Sciences and Engineering, 2009, Pages 1031
-
1035.

5.

Jixin Ma, “State
-
based
Time Series Analysis and Prediction

, IEEE, 2007.

6.

L. M.
Floria, C.
Gracia
-
Lazaro, “Social Network Reciprocity as a Phase Transition in Evolutionary
Cooperation”, IEEE, 2008.

7.

Sukeshini A. Grandhi, “Answer or Not to Answer: That is the question for cell phone user ”, ACM, 2009.

8.

Keivan Kianmehr, Reda Alhajj, “Effe
ctiveness of Machine Learning Technique for Automated
Identification of Calling Communities”, IEEE, 12
th

International Conference Information Visualization,
2008, Pages 308
-
313.

9.

Huiqi Zhang, Ram Dantu, “Quantifying
the Presence of Phone Use

, IEEE, 2008.

10.

Azlinah Mohamed, Ahmad Faud Mohamed Bandi, Abdul Razif Tamir, Md Daud Jaafar, Suriah Hasan,
Faeizah Jusof, “Telecommunication Fraud Prediction Using Backpropagation Neural Network”, IEEE,
International Conference of Soft Computing and Pattern Recognition,
2009, Pages 259
-
264.

11.

Chaogui Kang, Song Gao, Xing Lin, Yu Xiao, Yihong Yuan, Yu Liu, Xiujun Ma, “Analyzing and Geo
-
visualizing Individual Human Mobility Patterns Using Mobile Call Records ”.

12.

Chih
-
Chieh Hung, Wen
-
Chih Peng, “A Regression
-
Based Approach for
Mining User Movement Patterns
form Random Sample Data”, Elsevier, Data & Knowledge Engineering 2011, Pages 1
-
20.

13.

Lian Yan, Michael Fassino, Patrick Baldasare, “Predicting Customer Behavior via Calling Links”, IEEE,
Proceedings of International Joint Confer
ence on Neural Networks, Montreal, Canada, July 31
-
August 4,
2005, Pages 2555
-
2560.

14.

Chen Zhou, Zhengguang Xu, Benxiong

Huang, “Activity Recognition from Call Detail Record: Relation
Between Mobile Behavior Patterns and Social Attribute Using Hierarchical Conditional Random Fields”,
IEEE/ACM, International Conference on Green Computing and Communications & IEEE/ACM
Interna
tional Conference on Cyber, Physical and Social Computing, 2010, Pages 605
-
611.

15.

[Sherif Rashad, Mehmed Kantar
dzic, Anup Kumar, “User mobility oriented predictive call admission
control and resource reservation for next generation
mobile network

, Elsevier,

2006.

16.

Chih
-
Ping Wei, I
-
Tang Chiu, “Turning telecommunication call details

to churn prediction: a data mining
approach
”,

Elsevier, Expert Systems with Applications, 2002, Pages 103
-
112.

17.

Fei Yu, Victor Leung, “Mobility
-
based predictive call admission

contro
l and bandwidth reservation in
wireless cellular network

,

Elsevier, 201
1.

18.

Olga Brazhnik, John F. Jones, “Anatomy of data intergration”
, Elsevier, 2001.

19.

Xiao
-
Sheng Si, Chang
-
Hua Hu, Jian
-
Bo Yang, “On the dynamic evidential reasoning algorithm for fault
pre
diction”, Elsevier
, 2011.

20.

Hui Li, Young
-
Chan Lee, Yan
-
Chan Zhou, Jie Sun, “The random subspace binary logit (RSBL) model for
bankruptcy prediction”, Elsevier, Knowledge
-
Based Systems, 2011, pages 1380
-
1388.

21.

Anthony J.T.Lee, Yao
-
Te Wang, “Efficient Data Min
ing for Calling Path Patterns in GSM Networks”,
Elsevier, Information Systems, 2003, Pages 929
-
948.

22.

Keivan Kianmehr, Reda Alhajj, “Calling communities analysis and identification using machine learning
techniques”, Elsevier, 2008.

23.

Shin
-
Yuan, “Applying data

mining to telecom churn management”,
Elsevier, 2006.

24.

Mukund Seshadri, Sridhar Machiraju, “Mobile Call Graph: Beyond Power Law and Lognormal
distributions”, ACM, 2008.

25.

Nithi, Lipika Dey, “Anomaly Detection

from Call Data Record

, ACM, 2009.

26.

Azlinah

Mohamed, Ahmed Faud Mohamed Bandi, “Telecommunication Fraud Prediction using Back
-
propagation Neural Network”
, ACM, 2009.

27.

Sandra Tartarelli, Nico D’Heureuse
, “Lessons learned on the usage of call logs for security and
management in the IP telephony”, ACM
/IEEE, 2010.

28.

Seungjae Shin, Gilju Park, Wonjun Lee, Sunmi Lee, “How to Make Telecom Pricing Strategy Using Data
Warehouse Approach”, IEEE, 31
st

Annual Hawaii International Conference on System Sciences, 1998,
Pages 55
-
60.

29.

Hui Zang, Francois Baccelli, Jean
Bolot, “Bayesian Inference for Localization in Cellular Networks”,
IEEE, INFOCOM, 2010.

30.

Tom Fawcett, Foster Provost, “Adaptive Fraud Detection”, ACM, Data Mining and Knowledge Discovery,
1997, Pages 291
-
316.

31.

Anna A. Shaverdian, Hao Zhou, George Michailidis
, “Algebraic Visual Analysis: The Catalaon Phone Call
Data Set Case Study”, ACM, 2009, Pages 74
-
82.

32.

Giuseppe Bianchi, Nico dOHeureuse, Saverio Niccolini, “Public Review for On
-
demand Time
-
decaying
Bloom Filters for Telemarketer Detection”, ACM SIGCOMM Comp
uter Communication Review, Volume
4
1, November 5, October 2011
.

33.

Ramaswamy Hariharan, Ji Meng Loh, James Shanahan, Kenji Yamada, “Spatial probabilistic modeling of
calls to businesses”, ACM, Proceedings of the 18
th

SIGSPATIAL International Conference on Adv
ances in
Geographic Information Systems, 2010, Pages 466
-
469.

34.

Tain Zhu, Bai Wang, “Role defining user behavior based clustering in telecommunication network”, ACM,
2011.

35.

Muhammad Usman Khan, Shoab Ahmed Khan, “Social Network Identification and Analysis usi
ng Call
Detail Records”, ACM, Proceedings of the 2
nd

International Conference on Interaction Sciences:
Information Technology, Culture and Human, 2009, Pages 192
-
196.

36.

Umang Sharan, Jennifer Neville, “Exploiting Time
-
Varying Relationships in Statistical
Relational
Models”, ACM, 2007.

37.

David T. Kao, R. Daniel Bergeron, Ted M. Sparr, “Efficient Proximity Search in Multivariate Data”
, IEEE.

38.

Anqin Zhang, Wenbin Zhong, Weihua LIU, “Detection of Outlier Patterns in Call Records Based on
Skeleton Points”, IEEE, I
nternational Conference on Artificial Intelligence and Computing Intelli
gence,
2010, Pages 145
-
149.

39.

Fosca Giannotti, Micro Nanni, Dino Pedreschi, Fabio Pinelli, Chiara Renso, Salvatore Rinzivillo, Roberto
Trasarti, “Unveilling the Complexity of Human Mobil
ity by Querying and Mining Massive Trajectory Data
”, Springer, September 2010, Pages 695
-
719.

40.

Hui Zang, Jean C. Bolot, “Mining Call and Mobility Data to improve Paging Efficiency in Cellular
Networks”, ACM, 2007.

41.

Nathan Froyd, John Mellor
-
Crummey, Rob Fo
wler, “Low
-
Overhead Call Path Profiling of unmodified
optimized code”, ACM, 2005.

42.

Robert Mintram, Chukwudi Anyakoha
, “Modeling call detail records forms mobile telecommunication
network”, ACM, 2007.

43.

Keivan Kianmehr, Reda Alhajj, “Calling Communities Analys
is and Identification using Machine Learning
Technique”, Elsevier, Expert Systems with Appications, Volume 36, Issue 3, Part 2, April 2009, Pages
6218
-
6226.

44.

Fang Cheng, Lei Tian, Jinfeng Xie, “Research on Path ID and VPI/VCI relating guess solution for TD
-
SCDMA network centralized monitoring system ”, IEEE, Seventh International Conference on Intelligent
Information Hiding and Multimedia Signal Processing, 2011, Pages 395
-
398.

45.

Khairul Nizam Baharim, Mohd. Shafri Kamaruddin, Faeizah Jusof
, “Leveraging Missing Values in Call
Detail Record Using Naïve Bayes for Fraud Analysis”, Telekom Malaysia Berhad, Project No R06
-
0651
-
0.