LOGMIN: A MODEL FOR CALL LOG MINING IN MOBILE DEVICES

International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013
DOI:10.5121/ijfcst.2013.3405 39
LOGMIN: A MODEL FOR CALL LOG MINING IN
MOBILE DEVICES
K.S. Kuppusamy1
, Leena Mary Francis2
, G. Aghila3
1,3
Department of Computer Science, School of Engineering and Technology,
Pondicherry University, Pondicherry, India
1
kskuppu@gmail.com, 2
rosebeauty02@gmail.com, 3
aghilaa@gmail.com
ABSTRACT
In today’s instant communication era, mobile phones play an important role in the efficient communication
with respect to both individual and official communication strata. With the drastic explosion in the quantity
of calls received and made, there is a need for analyses of patterns in these call logs to assist the user of the
mobile device in the optimal utilization. This paper proposes a model termed “LogMin” (Log Mining of
Calls in Mobile devices) which is aimed towards mining of call log in mobile phones to discover patterns
and keep the user informed about the trends in the log. The logging of calls would facilitate the user to get
an insight into patterns based on the six different parameters identified by the proposed LogMin model. The
proposed model is validated with a prototype implementation in the Android platform and various
experiments were conducted on it. The results of the experiments in the LogMin Android implementation
validate the efficiency of the proposed model with respect to user’s relevancy metric which is computed as
96.52%.
KEYWORDS
Mobile Computing, Call Log Mining, LogMin
1. INTRODUCTION
Mobile phones have become the “numero uno” devices not only among the nerds but also with
general masses which is validated with the terrific explosion in the quantity of mobile devices.
The availability of more than 6.8 billion mobile devices across the globe emphasises the utility
and the necessity of these devices across the spectrum of people. The drastic increase in the
number of mobile devices is substantiated by the fact that the quantity of devices in 2010 is 5.4
billion and 6.1 billion in 2011. [1] With respect to the Indian context the total number of mobile
phone subscription is about 906 million amounting to 73% of the population which is in line with
the global average.
In addition to the drastic increase in the quantity of the mobile devices, their individual utility is
also on a steep rise. The utility of mobile devices among the youth in specific, itself has been
studied by researchers in various dimensions including the psychological analysis. [2], [3] As
with any other technology, the mobile phones do carry some adverse effects like behavioural
addiction. In order to handle these problems the study of mining the call log of mobile phones
becomes an interesting and necessary research issue which need to be addressed.
The general utility of mobile phones among the end users in terms of various aspects like
communication pattern, management of activities etc have been studied by researchers to
facilitate efficient handling of these mobile devices.[4].

40
This paper proposes a model titled “LogMin” (Log Mining of calls in mobile devices). The
objectives of the proposed LogMin model are as listed below:
 Proposing a model for mining the log of calls in the mobile devices to infer the user
relevant information from them.
 Providing a detailed insight into the call logs of the mobile phones to the users, in terms
of multiple components like Time-of-Day, Provider based grouping with the help of
clustering techniques.
The remainder of this paper is organized as follows: In Section 2, some of the related works
carried out in this domain are explored. Section 3 deals with the proposed model’s mathematical
representation. Section 4 is about prototype implementation and experiments. Section 5 focuses
on the conclusions and future directions for this research work.
2. RELATED WORKS
This section walks though some of the related works which have been carried out in this domain.
The proposed LogMin model attempts to explore the patterns in the call log and provide the user
various views along different dimensions. The mobile phones and their utility has been an
important research area in the information communication domain which has been studied by
various researchers to address specific issues.
The mobile phone logs have been studied by researchers for various tasks like clustering of users
and interface optimization purpose [5]. In addition to the logs, various other sensing abilities from
the mobile phones have been utilized by researchers in understanding the optimal utilization of
these devices [6] [7].
With respect to handling the call log of mobile phone, there are studies which have focused on
providing calendar specific categorization of logs [8]. There exist applications for various
smartphone platforms like Android for managing the call log in a professional manner which
provide facilities like extraction of log information in a specific format, providing the statistical
data regarding the log in the mobile devices [9]. These kind of applications provide various
services like rendering of log in specific manners apart from the traditional linear list manner,
facility to backup and restore the log information and the ability to manipulate the log information
at the lowest possible abstraction [10]. There are studies on reality mining based on the real time
datasets which considers various log information provided by the mobile devices [11]. In addition
to the individual log mining there are studies conducted at the corporate level to analyse the call
log of technical support etc [12]. There are studies conducted on the call logs from central
repositories with millions of calls [13]. The call pattern analysis is carried out using neural
networks by researchers [14].
The proposed LogMin model provides the call log results in the form of clusters based on
parameters like duration etc. There clustering techniques have been studied in detail in the
information management domain by various researchers [15], [16], [17]. The clustering
techniques have been adopted for various applications like image handling, web page
manipulation etc [18]. The proposed model utilizes the k-Means clustering technique for the
purpose of clustering the call logs into various groups based on different parameters. The k-
Means clustering algorithm allows to cluster a dataset into k-clusters and it belongs to the
supervised machine learning category [19], [20]. The reasons for selecting the k-Means algorithm
are the ability to cluster faster than other techniques like hierarchical, fuzzy clustering and the
ability to produce the clusters in a cohesive manner.

41
The proposed LogMin model extracts the call log information from the mobile devices and the
performs the analysis on the log data based on various parameters and provide the clustered result
set with the help of k-Means clustering technique. The details on the model are illustrated in the
following section.
3. THE LOGMIN MODEL
This section explores about the building blocks of proposed LogMin model for analysing the call
log of the mobile phones and inferring details from it. The block diagram of the proposed LogMin
model is as shown in Fig. 1.
Figure 1. LogMin – Block Diagram
The components of the proposed LogMin model are as explained below:
 Call Log Fetcher: The role of Call Log Fetcher is to fetch call log information from the
device’s call repository.

42
 Call Log Parser: The Call Log Parser receives the input from the call log fetcher and its
primary aim is to parse the input and store it appropriate locations. It has got the
following sub components.
o Log Item Structure: The log item structure is a data structure used to hold
component data of the call log. This structure has following fields: The “Call Log
Id” which is a unique number assigned to every single entry in the log in order to
identify a call item uniquely. This is assigned by the model automatically in an
auto-increment manner. The “Call Number” is the phone number of the device
from which the call has originated. In case of outgoing calls it would hold the
value of “SELF” which indicates that the caller is the current device. In case of
items in the contact there would be a “Caller Name” field. In case of new number
this field would be populated as “Un-Known”. The next field in the structure is
the “Start-Time” of the Call and the last field is the “End-Time” of the call. This
structure’s visual representation is as shown in Figure 2.
Figure 2: Log Item Structure
o Log Store: Another sub-component in the Log parser is the “Log Store” which
has the facility to store the calls into three different sectors which are termed as
“Incoming (IC) Store”, “Outgoing (OC) Store” and “Missed Call (MC) Store”.
The missed calls are identified by comparing the start time and end time in the
call log structure. In cases where they are equal that particular call is identified as
the missed call.
 Log Miner: The primary role of Log Miner is to mine the logs in all the three different
stores explained above. The mining of the calls is carried out using six different
parameters which are identified as “ToD – Time of the Day”, “Duration”, “Frequency”,
“Provider”, “Boundary”, “CC -Conference Count”. The boundary parameter indicates the
calls which the user has either made or received in both sides of the time dimension. The
“Conference Count” is another parameter which indicates the parallel calls i.e. the
conference calls made with the current call. The concurrent calls are identified with the
the help of time window which is obtained by comparing the start time and end time of
these calls.
 LogMin Reporter: The LogMin reporter is component is responsible for providing the
reports based on the results of Mining component. The various reports which are
presented to the user are Frequency report, calendar report, ToD (Time of Day) report and
“MNP – Mobile Number Portability report”. These reports are generated by receiving the
inputs from the miner component. The frequency report is provided in two dimensions
viz., Count and Cluster. The cluster report is generated by clustering the calls based on
the frequency which is done with the help of k-Means clustering algorithm. The calendar
report is presented in three dimensions viz., “MV – Month View”, “WV – Week View”
and “DV Day View”. Another report presented to the user is the “TOD report” which is
again rendered in three different variations viz., “MR- Morning Report”, “AR –
Afternoon report” and “ER – Evening Report”. These are based on the Time of Day at

43
which the calls are either received or made. Another report generated by the LogMin
reporter is the mobile number portability report which is based on optimizing the tariff to
the user based the call log i.e. if the user is making frequent calls numbers belonging to
the a particular provider other than the current provider then the decision to port the
number to that service provider is analysed. The portability index is computed and if it
goes beyond the threshold value then MNP is advised by the LogMin.
3.1 Mathematical Model
This section explores the mathematical representation of the proposed LogMin model. The
LogMin call log store with its three components is represented as shown in (1).



 
 
   
 
 
(1)
In (1),  indicates the Call Log Store,  indicates the IC Store,  indicates the OC Store and
 represents the MC Store. The Log Structure is represented as shown in (2).
G





 
 
  
   
 
 
  
(2)
In (2), G is the generic Call Log Structure, and the five components indicated in Figure 2 are
represented from  to  respectively.
The generic G is extended to all the three specific log stores as shown in (3).
IC
OC
MC
 
 
   
  
(3)
The entries in all these stores are analysed using six different dimensions as indicated in (4).

44
M







 
 
 
  
  
 
 
 
  
(4)
In (4), M indicates the mining of IC Store and six different parameters utilized for the purpose of
mining are represented from  to  respectively. The parameters are computed individually. The
computation of ToD parameter  is as shown in (5). The ToD is computed as the mean of start
time and end time which are represented as  and  respectively.
2
M
 






 
 
 
 
 
  
 
 
 
 
 
(5)
The duration parameter  is computed as shown in (6). The duration is computed as the difference
between the start and end time.
2
M
 

  




 
 
 
  
 
  
 
 
 
 
 
(6)
The frequency parameter is computed as the count of calls in the reference time as shown in (7).

45
2
( ) : ( ) RM T
 

  
  



 
 
 
  
      
 
 
 
 
 
(7)
In (7), RT indicates the Reference Time after which all the computation is made. The reference
time is incorporated to avoid very old calls skewing the results. The Provider data is retrieved
from the online repositories as shown in (8).
  
2
( ) : ( ) RM T
 

  
  
 


 
 
 
  
      
   
 
 
 
 
(8)
The boundary parameter is computed as the set of entries in the call log before and after the call
with in a reference time, as shown in (9).
  
 
2
( ) : ( )
( ) : ( )
R
M
F P
T
T T
 

  
  
 
   

 
 
 
  
      
   
 
     
 
 
(9)
In (9), FT indicates the Reference time for calls after the current one, and PT indicates the
Reference time for calls before the current one. The conference count parameter is computed by
analysing the start and end time of calls and their proximity, as the conference calls would have
near equal, start time and end time. The computation is as shown in (10).

46
  
 
2
( ) : ( )
( ) : ( )
( ) : ( ); ( )
R
M
F P
T
T T
 

  
  
 
   
     
 
 
 
  
      
   
 
     
 
      
(10)
The same procedure is repeated with corresponding data for M and M . After the completion of
parameter computation, the LogMin miner analyses the data with statistical techniques for some
of the reports and it utilizes the k-Means clustering technique for computing the clusters in the log
data. The clustering process is as indicated in (11).
   1 2, ..
0,
i i k
CLM
i
        
  
      
(11)
In (11), CLM indicates LogMin Cluster,  indicates the k-Means clustering engine, and each
i represents a cluster, and “k” indicates the cluster size. The result of this process is given to the
LogMin reporter which generates the reports as per the user’s request.
4. EXPERIMENTS AND RESULT ANALYSIS
This section explores the experimentation and results associated with the proposed LogMin
model for mining the call log of mobile devices. The prototype implementation is done with the
Android platform [21]. The prototype application is tested with an array of hardware running
Android operating system. Though the prototype is made with the Android platform, the same
shall be extended to other environments like iOS and Windows Mobile. The screenshot of the
application is shown Figure 3.

47
Figure 3: Screenshots of LogMin Android Application
The experiments were conducted with 15 groups of users covering a spectrum from Novice,
General to Geek users. The relevance of the data mined with the LogMin model was analysed
with these groups of users and the experiments were conducted in various sessions. The results of
the experiments are illustrated in Figure 4 and Table 1. During the experiments the users were
asked to mark the information rendered by the LogMin reporter into three different categories
viz., “CRI – Completely Relevant Information”, “PRI – Partially Relevant Information” and “CII
– Completely Irrelevant Information”.
Table 1: LogMin Information Relevance
Session
ID
CRI PRI CII
1 88.4 6.4 5.2
2 82.6 10.3 7.1
3 85.5 12.8 1.7
4 86.2 12.5 1.3
5 88.1 11.4 0.5
6 80.1 14.5 5.4
7 82.1 10.5 7.4
8 88.5 6.5 5
9 90.2 6.5 3.3
10 92.3 5.5 2.2
11 95.1 3.2 1.7
12 94.2 4.5 1.3
13 91.1 5.9 3
14 90.1 6.5 3.4
15 89.8 6.5 3.7

48
Figure 4: LogMin Relevance Comparison
Figure 5: LogMin CRI vs PRI vs CII
It can be observed from the experimental results that the mean of CRI across the sessions is
88.2%, the mean of PRI as 8.23%. The CII component is observed only at a negligible level of
3.48% as illustrated in Figure 5. As the cumulative of CRI and PRI is evaluated as 96.52% the
efficiency of the LogMin model in providing the relevant information from the call log is
confirmed.

49
5. CONCLUSIONS AND FUTURE DIRECTIONS
The following are the conclusions derived from the LogMin model:
 LogMin model provides a detailed insight into the calling behaviour of the mobile user
which shall be used to improve the effectiveness of utilization.
 The proposed LogMin model facilitates the retrieval of user relevant information from
the logs of calls from the mobile device.
 The efficiency of the LogMin model is confirmed with the user relevance metric with a
cumulative PRI and CRI value of 96.52%.
The future directions for the LogMin model include the following:
 The LogMin model shall be extended to include the log information other than the call
log like GPS and other sensor data.
 The model shall be further enriched by incorporating collaborative call log mining which
would include the log of users who communicate with each other frequently, thereby
forming call log based network.
REFERENCES
[1] “Global mobile statistics 2013 Part A: Mobile subscribers; handset market share; mobile operators.”
[Online]. Available: http://mobithinking.com/mobile-marketing-tools/latest-mobile-
stats/a#subscribers. [Accessed: 27-May-2013].
[2] S. P. Walsh, K. M. White, and R. M. Young, “Over-connected? A qualitative exploration of the
relationship between Australian youth and their mobile phones,” Journal of Adolescence, vol. 31, no.
1, pp. 77–92, Feb. 2008.
[3] C. O. Güzeller and T. Coşguner, “Development of a Problematic Mobile Phone Use Scale for Turkish
Adolescents,” Cyberpsychology, Behavior, and Social Networking, vol. 15, no. 4, pp. 205–211, Apr.
2012.
[4] “Fast Capitalism.” [Online]. Available:
http://www.uta.edu/huma/agger/fastcapitalism/1_1/rippin.html. [Accessed: 27-May-2013].
[5] M. H. Jeon, D. Y. Na, J. H. Ahn, and J. Y. Hong, “User segmentation & UI optimization through
mobile phone log analysis,” in Proceedings of the 10th international conference on Human computer
interaction with mobile devices and services, New York, NY, USA, 2008, pp. 495–496.
[6] N. D. Lane, E. Miluzzo, H. Lu, D. Peebles, T. Choudhury, and A. T. Campbell, “A survey of mobile
phone sensing,” IEEE Communications Magazine, vol. 48, no. 9, pp. 140–150, 2010.
[7] Y. Lee and S.-B. Cho, “Extracting Meaningful Contexts from Mobile Life Log,” in Intelligent Data
Engineering and Automated Learning - IDEAL 2007, H. Yin, P. Tino, E. Corchado, W. Byrne, and X.
Yao, Eds. Springer Berlin Heidelberg, 2007, pp. 750–759.
[8] “Call Log Calendar - Android Apps on Google Play.” [Online]. Available:
https://play.google.com/store/apps/details?id=jp.joao.android.CallLogCalendar&feature=search_resul
t. [Accessed: 27-May-2013].
[9] “Call Log Manager Pro - Android Apps on Google Play.” [Online]. Available:
https://play.google.com/store/apps/details?id=com.appsbeyond.android.callhistoryplus&feature=searc
h_result. [Accessed: 27-May-2013].
[10] [10]“Call Log + - Android Apps on Google Play.” [Online]. Available:
https://play.google.com/store/apps/details?id=com.contapps.android.call_log&feature=search_result.
[Accessed: 27-May-2013].
[11] N. Eagle and A. (Sandy) Pentland, “Reality mining: sensing complex social systems,” Personal
Ubiquitous Comput., vol. 10, no. 4, pp. 255–268, Mar. 2006.
[12] “Pragmatic Text Mining: Minimizing Human Effort to Quantify Many Issues in Call Logs - HPL-
2006-60R1.pdf.” [Online]. Available: http://www.hpl.hp.com/techreports/2006/HPL-2006-60R1.pdf.
[Accessed: 27-May-2013].
[13] M. Seshadri, S. Machiraju, A. Sridharan, J. Bolot, C. Faloutsos, and J. Leskove, “Mobile call graphs:
beyond power-law and lognormal distributions,” in Proceedings of the 14th ACM SIGKDD
international conference on Knowledge discovery and data mining, New York, NY, USA, 2008, pp.
596–604.

50
[14] “Data mining, fraud detection and mobile telecommunications... - etd_init_3937_1174040706.pdf.”
[Online]. Available: http://etd.uwc.ac.za/usrfiles/modules/etd/docs/etd_init_3937_1174040706.pdf.
[Accessed: 08-Jul-2013].
[15] S. Guha, R. Rastogi, and K. Shim, “CURE: an efficient clustering algorithm for large databases,” in
Proceedings of the 1998 ACM SIGMOD international conference on Management of data, New
York, NY, USA, 1998, pp. 73–84.
[16] J. A. Hartigan, Clustering Algorithms, 99th ed. New York, NY, USA: John Wiley & Sons, Inc.,
1975.
[17] J. A. Hartigan and M. A. Wong, “Algorithm AS 136: A K-Means Clustering Algorithm,” Applied
Statistics, vol. 28, no. 1, p. 100, 1979.
[18] S. Alcic and S. Conrad, “A clustering-based approach to web image context extraction,” in MMEDIA
2011, The Third International Conferences on Advances in Multimedia, 2011, pp. 74–79.
[19] T. Kanungo, D. M. Mount, N. S. Netanyahu, C. D. Piatko, R. Silverman, and A. Y. Wu, “An efficient
k-means clustering algorithm: analysis and implementation,” IEEE Transactions on Pattern Analysis
and Machine Intelligence, vol. 24, no. 7, pp. 881–892, 2002.
[20] C. Ding and X. He, “K-means clustering via principal component analysis,” in Proceedings of the
twenty-first international conference on Machine learning, New York, NY, USA, 2004, p. 29–.
[21] “Android.” [Online]. Available: http://www.android.com/. [Accessed: 28-May-2013].
Authors
Dr. K.S.Kuppusamy is an Assistant Professor at Department of Computer Science, School
of Engineering and Technology, Pondicherry University, Pondicherry, India. He has
obtained his Ph.D in Computer Science and Engineering from Pondicherry Central
University, and the Master degree in Computer Science and Information Technology from
Madurai Kamaraj University. His research interest includes Web Search Engines,
Semantic Web and Mobile Computing. He has made more than 20 peer reviewed
international publications. He is in the Editorial board of three International Peer Reviewed Journals.
Leena Mary Francis was associated with Oracle as Software Engineer before she joined as
Assistant Professor at Department of Computer Science, SS College, Pondicherry, India.
She has obtained her Master’s degree in Computer Applications from Pondicherry
University. Her area of interest includes Web 2.0 and Information retrieval.
Dr. G. Aghila is a Professor at Department of Computer Science, School of
Engineering and Technology, Pondicherry University, Pondicherry, India. She has got a
total of 24 years of teaching experience. She has received her M.E (Computer Science
and Engineering) and Ph.D. from Anna University, Chennai, India. She has published
more than 75 research papers in web crawlers, ontology based information retrieval.
She is currently a supervisor guiding 8 Ph.D. scholars. She was in receipt of Schrneiger
award. She is an expert in ontology development. Her area of interest includes
Intelligent Information Management, artificial intelligence, text mining and semantic web technologies.

LOGMIN: A MODEL FOR CALL LOG MINING IN MOBILE DEVICES

More Related Content

Viewers also liked (6)

Similar to LOGMIN: A MODEL FOR CALL LOG MINING IN MOBILE DEVICES (20)

More from ijfcstjournal (20)

Recently uploaded (20)

LOGMIN: A MODEL FOR CALL LOG MINING IN MOBILE DEVICES