Content uploaded by Shylaja .S
Author content
All content in this area was uploaded by Shylaja .S on Nov 08, 2017
Content may be subject to copyright.
IRACST – International Journal of Computer Networks and Wireless Communications (IJCNWC), ISSN: 2250-3501
Vol.7, No 3, May-June 2017
1
APPLICATION OF DATA MINING TECHNIQUES IN
WIRELESS SENSOR NETWORKS:
A REVIEW
S. Shylaja1
1Assistant Professor & Research Scholar,
Department of Computer Science,
Rathinam College of Arts & Science (Autonomous)
Coimbatore-641021, Tamil Nadu, India,
shylaja.cs@rathinamcollege.com
R. Muralidharan2
2 Vice Principal & Head,
Department of Computer Science,
Rathinam College of Arts & Science (Autonomous),
Coimbatore-641021, Tamil Nadu, India,
hod.cs@rathinamcollege.com
Abstract - Wireless Sensor Network is one of the most important
technologies, which is being used in computer science. It consists of
smart sensor nodes which can sense and communicative in a wireless
fashion in a defined a network and it get the data to sink. The sink is
connected to gateway and collected data’s are sent to the base station.
Sensors have many constraints like limited resources, energy,
memory, computation power etc. Data Mining is used to evaluate
data from several dimensions and is the process of finding pattern in
a large relational database. The purpose of using data mining is to
increase the efficiency of the sensor nodes. The main objective of this
review focused on distinct data mining techniques which have been
adopted by WSN showing their advantages and disadvantages with
various types of applications.
Keywords - Data Mining Techniques, Wireless sensor networks,
Centralized approach, Distributed approach
I INTRODUCTION
A. Wireless Sensor Networks: Advantages and
Disadvantages
In Today’s scenario wireless sensor network has become very
essential in our daily uses, it is considered as one of the most
important technology for the twenty-first century without
WSN our work would have been very difficult. WSN can
reach into those areas where human cannot even think to reach
like disaster areas, space, sea etc. WSN has earned terrific
attention from both industry and academic all over the world.
Wireless sensor network consist of thousands of low cost
nodes, low-power, and multifunctional senor that are deployed
in a region of interested area which run on battery. The sensor
nodes generally are small in size each node has one or more
embedded microcontroller i.e. CPU or DSP chip set. These
nodes have sensing capabilities along with communication
capabilities. These nodes interact over a short area through a
wireless medium which are organized into ad-hock network
[7].
Advantages and Disadvantages of WSN
Why people love wireless sensor networks might be
summarized as the following [12]:
• Network setups can be carried out without fixed
infrastructure.
• Suitable for the non-reachable places such as over the
sea, mountains, rural areas or deep forests.
• Flexible if there is random situation when additional
workstation is needed.
• Implementation pricing is cheap.
• It avoids plenty of wiring.
• It might accommodate new devices at any time.
• It's flexible to undergo physical partitions.
• It can be accessed by using a centralized monitor.
The disadvantages of wireless sensor networks can be
summarized as follows [12]:
• Less secure because hackers can enter the access
point and obtain all the information.
• Lower speed as compared to a wired network.
• More complicated to configure compared to a wired
network.
• Easily troubled by surroundings (walls, microwave,
large distances due to signal attenuation, etc.).
• It is easy for hackers to hack it we couldn’t control
propagation of waves.
• Comparatively low speed of communication.
• Gets distracted by various elements like Blue-tooth.
• Still Costly (most importantly)
B. Applications of Wireless Sensor Networks
The Applications for WSNs involve tracking, Monitoring and
Controlling. WSNs are mainly utilized for habitat monitoring,
object tracking, nuclear reactor control, fire detection and
traffic monitoring.
There are lots of applications in wireless sensor networks [12]
• Process Management: Area monitoring is very
common using WSNs. In area monitoring, the WSN
is deployed spanning a region where some
phenomenon is usually to be monitored. A military
example may be the use of sensors detect enemy
intrusion; a civilian example would be the geo-
IRACST – International Journal of Computer Networks and Wireless Communications (IJCNWC), ISSN: 2250-3501
Vol.7, No 3, May-June 2017
2
fencing of gas. Area monitoring is most important
part.
• Healthcare monitoring: The medical applications
might be of two sorts:
9 Wearable
9 Implanted
Wearable devices are applied to the body surface of
the human or maybe at close proximity from the user.
The implantable medical devices are the ones that are
inserted inside your body. It is used to monitoring of
ill patients in hospitals and also at home.
• Environmental/Earth sensing: There are numerous
applications in monitoring environmental parameters
samples of which are given below. They share any
additional challenges of harsh environments and
reduced power supply.
• Polluting of the environment monitoring: Wireless
sensor networks have been deployed in lots of cities
(Stockholm, London and Brisbane) to monitor the
power of dangerous gases for citizens.
• Forest fire detection: A network of Sensor Nodes is
usually positioned in a forest to detect every time a
fire has begun. The nodes are usually with sensors to
measure temperature, humidity and gases which are
produced by fire within the trees or vegetation. The
first detection is necessary to get a successful action
of the fire fighters; As a result of Wireless as Sensor
Networks, the fire brigade are able to know when a
fire begins you bet it can be spreading.
• Water quality monitoring: Water quality monitoring
involves analysing water properties in dams, rivers,
lakes & oceans, and also underground water reserves.
All the applications which are mentioned above are reliable
and real-time monitoring is the essential requirements. All the
application generate huge volume of data which are fast
changing and heterogeneous in nature. All the data is collected
and filtered into useful information or pattern by using data
mining techniques. As in upcoming years the whole world will
be ruled by wireless sensors, so it will be very crucial to
choose a correct data mining technique which is a big
challenge in WSNs.
The pattern of paper is as follows: Section2 explains about
data mining technique. Section 3 explains about WSN in brief.
Section 3 deals with the literature survey and section 4
describe the conclusion of the whole paper.
II DATA MINING TECHNIQUES
The data mining is elucidated as heart of enlightenment
discovery process. It can be briefed as the process of gathering
facts from different area and refining it into appropriate
knowledge. It can also be defined as the evocation of hidden
predictive knowledge from a huge database. Technically, from
a large relational databases a search is performed among
several field areas to obtain a useful pattern which can be used
in future. The motive of using data mining is to help
companies to focus on the most important data in their data
warehouses. The main idea of data mining technique is to
extract data from large dataset and convert into some useful
pattern for future use. Data mining is divided into two models,
descriptive and predictive. All Data mining techniques fall
under these two categories [4].
• Predictive model - The primary goal of using this
exemplary is that we can predict the future result
than the current situation. It falls under the
supervised learning and the predicted output can be
numeric as well as in categorized form, as it always
predicts the target value.
• Descriptive model - This method is generally used to
generate correlation, frequencies, cross tabulation etc.
It is used to discover regularities in the data and
uncover patterns. From bulk of data, a search is
performed for finding interesting subgroup patterns.
Figure 1 : Data Mining Models
The main steps are for knowledge discovery stage is as
follows:
1. Data Cleaning: It is the first step in which
inconsistent data and noises are removed.
2. Data Integration: It is the second step where the
combination of multiple data source is done.
3. Data Selection: It is the third step where
necessary data are searched from the database to
reach the goal.
4. Data Transformation: It is also called as data
consolidation. In this stage the selected e data are
transformed into useful pattern for mining
5. Data mining: It is the most crucial stage in
which the data mining techniques are applied on
transformed data so that the tendency of the
extracted pattern can be determined.
6. Pattern Evaluation: It is the sixth step in which
only the important pattern which holding
accurate information are chosen.
Knowledge Presentation: It is the final stage in which the
mined data are provided to the user using virtualization and
knowledge representation techniques
IRACST – International Journal of Computer Networks and Wireless Communications (IJCNWC), ISSN: 2250-3501
Vol.7, No 3, May-June 2017
3
Figure 2: Stages for Knowledge Discovery process
III DATA MINING IN WIRELESS SENSOR NETWORKS
Data mining in sensor network is the method of selecting
application oriented standards and patterns with acceptable
accuracy from a continuous, fast and probably non ended flow
of data streams from sensor networks. In this case, all data
cannot be stored and must be processed quickly, Data mining
method has to be sufficiently fast to process high speed
arriving data. The overall goal of the data mining process is to
extract information from a data set and transform it into an
understandable structure for further use [15]
There are four types of data mining techniques which are
commonly used in WSNs which uses both centralized and
distributed approach that is given below
Figure 3: List of Data mining techniques in Wireless
Sensor Networks
There are three essentials elements for data mining without
these nothing can be achieved,
They are as follows;
1. Data: are any facts, numbers, or text that can be processed
by a computer and these growing amounts of data in different
formats and different databases are stored. The data which
defines other data is called met data such as design of logical
database.
2. Information: Any data which provide information in form
of pattern or relationship.
3. Knowledge: The patterns which are formed or gained are
converted into knowledge for future use
IV WIRELESS SENSOR NETWORKS
A wireless sensor network (WSN), it is a wireless network
which dwell of partially disperse autonomous equipment’s that
uses sensors to monitor environmental and real condition. A
WSN is a collection of large sensor nodes which can be in
hundreds or even thousands of small, cheap nodes which are
deployed into a network at certain location. These sensors
have the capability to sense, process and communicate to its
peer in order to work together in a cooperative manner. A
sensor network consist of thousands of battery which sense
huge bulk of data and send it back to the Sink through gateway
for processing purpose by using different data mining
techniques. Each node consists of processing capability in
form of microcontrollers that is CPU or DSP chip sets. WSNs
have different endowment like denser level of node
deployment, sever energy, computation power, data generated
by sensor nodes are bulk in size, fast changing and have very
less resource constraints [3]. In figure shown below consist of
thousands of sensor node which are arranged in a desired
location, the jobs of senor nodes are to collect bulk of data and
transfer it to number of gateway. They get the data to a sink
and the sink is wired to a gateway over a network, all the data
are transfer to the base station where useful data are sorted for
future work. The nodes communicate with each other in a
wireless manner which follows ad-hoc network. At present
WSNs consist of thousands of nodes which in future may
contain more than millions of nodes. In order to solve huge
volume of data we need to solve all the major issues of the
sensors like processing capability, limitation in memory, etc.
Figure 4: Wireless Sensor networks
IRACST – International Journal of Computer Networks and Wireless Communications (IJCNWC), ISSN: 2250-3501
Vol.7, No 3, May-June 2017
4
Other issues these nodes are, in future we have to deal with
heterogeneous data like sound, smell, taste, image, audio,
video, etc. In this kind of data, significant amount of energy is
consumed. There are basically two types of approaches in
WSNs which are mentioned below:
Table 1 Difference between Distributed and Centralized
Sensor
Application Distributed Centralized
Energy Limited No Constraints
Data Length Un Bounded Bounded
Response time Real Time Non Real Time
Update Speed High low
Number of Passes Single Multi task
Data Flow Continuous Stationary
Data Type Dynamic Static
A. Centralized Approach
Generally the data mining technique uses centralized
approach. In this approach data are collected directly at a
central location which is not protected or restricted by any
computational resources. This approach involves the selection
of cluster head by the base station itself. It is done on the basis
of amount of residual energy and distance of sensor nodes
from base station. In today’s scenario the data which are
generated are very fast developing and are huge in volume as
the memory storage is limited. The main problem in this
approach is, all the data collected are transferred to the central
node and if the central node fails then the whole network will
fail. This approach is mostly inappropriate due to scalability
issue like limited communication bandwidth, power supply
and less storage space.
B. Distributed Approach
This approach is better than centralized approach as the data is
not collected at a central node, if one node fails it will not have
any effect on any other nodes. In this method the resources are
need to be self-organized. The reliability of senor nodes can be
improved by the inducement of distributed architecture. This
approach helps in saving energy up-to a great extent. This
technique is used in order to control the issues like:-[7]
• Sensor nodes are prone to failure
• Better collection of data at sensor nodes
• To provide backup with nodes if failure
occurs in the central node.
V RELATED WORKS
Zahra haihashemi and mihail [1] proposed the A
multidimensional Time Series similarity measure with
application to eldercare monitoring. Here, data mining
techniques have been applied to sensor data in a wide range of
application domains such as Health monitoring systems,
manufacturing processes, Intrusion detection, database
management and others. Many data mining techniques are
based on computing the similarity between two sensor data
patterns. Then our author described a novel method for
computing the similarity of two multi attribute time based on
temporal version of smith waterman (TSW-Temporal smith
waterman) is a well-known bio informatics algorithm. This
method was used to sensor data from eldercare applications
for early illness detection. And these types of methods
mitigate difficulties related to data uncertainty and aggregation
that often arise when processing sensor data. Finally, The
WTSW (Window based temporal smith waterman) could be
potentially slow for eldercare applications. So they propose a
genetic version of it GATSW (Genetic algorithm with TSW)
and they tested our algorithm on multiple datasets.
Marjorie skubic et.al [2] presents an example of
unobtrusive, continuous monitoring in the home for the
purpose of accessing early health changes. Sensors embedded
in the environment capture behaviors and activity patterns. A
one dimensional (1-D) alter algorithm was used to generate
health alerts to clinicians in a senior housing facility. Here,
there were four classification approaches that use multi sensor
data. Results are shown using embedded sensor data and
health alert ratings collected on 21 sensors over nine months.
In this paper they were used four classifier techniques: Fuzzy
pattern tree (FPT), Fuzzy K-Nearest Neighbor (FKNN),
Neural Networks (NN), and Support Vector Machine. Those
multi-dimensional classifiers performed significantly better
than the 1-D algorithm, with the best 6D performance at 86%
compared to the 39% at 1D. And the 6D classifier based only
on the Domain Knowledge performed better than the best 6D
classifier using supervised machine learning.
Azhar Mahmood [3] has proposed the survey for
Mining Data Generated by Sensor Network. The Sensor
Network’s (SNs) produces huge accounts of data which offer
promising prospect for the application of data analysis
techniques to extract useful information for the end user. The
DM community had observed that extracting knowledge from
SNs Data through two methods. Network side processing
technique, that requires real time analysis methodology and to
handle dynamic data streams or events. Centralized processing
through high end computing is required for generating off-line
predictive insights which facilitates real-time analysis. From
analysis, it was observed that the techniques intended for
mining sensor data at N/W side are helpful for taking real time
decision as well as serve as prerequisite for development of
effective mechanism for Data storage, retrieval, query and
transaction processing at the Central Side.
D.J.Dechnene [4] had conducted a survey of
clustering algorithm for Wireless Sensor Networks. They
described the improvements to be made in the clustering
algorithm for wireless sensor networks. They suggested the
following three types of schemes Heuristic Scheme,
Weighted Schemes, Grid Schemes
Marsinisso Saoudi [5] had described the Data mining
Techniques that are applied to Wireless Sensor Networks for
Early Forest Fire Detection. They proposed a new approach
for forest fire detection based on the integration of data mining
techniques into sensor nodes. The idea is to use CLUSTERED
WSN.
IRACST – International Journal of Computer Networks and Wireless Communications (IJCNWC), ISSN: 2250-3501
Vol.7, No 3, May-June 2017
5
VI CONCLUSION
In this paper, we have explained different data mining method
which has been adopted for wireless sensor network which
uses different classification and evaluation approaches. We
inspected that each data mining technique mentioned here
works on either centralized or distributed approach are
discussed. However the main consideration of the approach
was information extraction and analysis on data with respect
to WSNs, We have analyzed and briefly described both
advantages and disadvantages of each algorithm. The entire
algorithm which has been discussed above is capable of
solving different types of issues related to WSNs types and
applications. The major challenges will be related to hardware
like compressing, storing and filtering of huge volume of data
generated by sensor nodes. Also in future we have to deal with
heterogeneous kinds of data like images, sounds, smell,
location and etc. The main motive to present this paper is
summarize, analyze and extract knowledge from data into
useful ideas for better decision making.
REFERENCES
[1] Zheng and jamalipour.A, “A wireless sensor networks: A networking
prospective”, John Wiley & sons, 2009.
[2] Zhao, X. and Sayed, A. H. (2012). Performance limits for distributed
estimation over lms adaptive networks. Signal Processing, IEEE Transactions
on, 60(10):5107–5124.
[3] A.R Raut and S.P Khandait, “ Review on Data Mining Techniques in
wireless sensornetworks” Electronics and communication systems (ICECS)
2015, 2nd International conference on Coimbatore 2015, PP 390-394.
[4] Pradnya P. Soundwale, “Overview of predictive and descriptive Data
Mining techniques”, IJARCSSE, Vol-5, 2015, PP 262-265.
[5] Bouyer, A. Hatamlou,A and Masdari,M “ A new approach for decreasing
energy in wireless sensor networks with hybrid LEACH protocol and fuzzy c-
means algorithm” International journal of communication network and
distributed system, 15(4), 2015, PP 400-412
[6] Shrikant D. Dhamdhere, Shanthi K, Guru “ Robust Data collection in
wireless sensor networks with mobile sink”, International journal of Computer
Science and Information technologies, vol 5 (4), 2014, PP 4999-5002.
[7] Aditya kumar naik, Rajendra kumar, Dwivedi “ A Review on use of Data
Mining methods in wireless sensor networks” ISSN ( Print) 2393-8374,
Online: 2394-0697, Vol 3, Issues 12, 2016, International Journal of current
engineering and scientific research (IJCESR).
[8] Massinissa saoudi, Ahcene Bounceur, Reinhardt Euler: Data Mining
techniques applied to wireless sensor networks for early forest fire detection”
ISBN 978-1-4503-4063-2. ICC’16, March 2016, Cambridge, United Kingdom
ACM.
[9] Mirjana Maksimovic, Vladimir Vujovic “ Comparative analysis of Data
Mining techniques applied wireless sensor networks data for dire detection”
JIJT3 (2013)2:65-77, Journal of information technology and application,
December 2013.
[10] Mohammed Alsheikh, Shaowei Zin, Dusit Niyato and Hwee-Pink Tan”
Machine learning in Wireless sensor network: Algorithm, Strategies and
Applications, arXiv: 1405, 4463v2 [cs.N1] 19 March 2015.
[11] Shubhie agarwal, Seema Matrey, Pooma Rana, Pankaj Singh Yadav “A
reasonable exploration of Data Mining techniques in WSN: International of
journal of Computer applications (0975-8887) vol 151- No 10, Oct 2016.
[12] Prashant Tiwari, Varren prakash saxena, Raj Gauravmishra, Devendra
Bhavsar, “Wireless sensor network: Introduction, advantages, applications and
research challenges” HCTL open international journal of technology
innovation and research (IJTIR) Vol 14, April 2015, ISBN 978-1-6-2951-946-
3.
[13] Marjorie skubic, Rainer Dane Guevara and Marilyn rantz “Automated
health alerts using in home sensor data for embedded health assessment” IEEE
journal of transactional engineering in health and medicine. Received 29 Oct
2014, Revised 28 Dec 2014, Accepted 6 mar 2015, Date of publication 10 Apr
2015. Date of current version 22 Apr 2015.
[14] Azhar Mahmood, Ice Shi and Shaheer Khatoon “Mining Data Generated
by Sensor Network, A survey” Information Technology Journal II :
1532–1543, 2012 ISSN 1812-5638
[15] Sherin Mathew, ―A Novel Approach for Sensory Data Collection in
Wireless Sensor Networks with Mobile Sinks, International Journal of
Engineering Science and Innovative Technology (IJESIT), Volume 2, Issue 2,
March 2013
[16] Shrikant D. Dhamdhere, ―Robust Data Collection in Wireless Sensor
Networks With Mobile Sinks, International Journal of Computer Science
and Information Technologies, Vol. 5 (4), 2014
[17] S.Punitha Shankari -Energy Saving In Wireless Sensor Network Using
Burst Algorithm, IOSR Journal of Electronics and Communication
Engineering (IOSR-JECE), Volume 5, Issue 1 (Jan. Feb. 2013)
[18] V.Karthik, -Region Based Scheduling With Multiple Mobile Robots for
Data Collection Strategies in Wireless Sensor Networks, International
Journal of Innovative Research in Science, Engineering and Technology, Vol.
2, Issue 7, July 2013
[19]Fan Bai, - A Novel Information Acquisition Technique for Mobile
Assisted Wireless Sensor Networks, IEEE Transactions on Vehicular
Technology, Vol. 61, No. 4, May 2012
[20] MiaoZhao, ―Bounded Relay Hop Mobile Data Gathering in Wireless
Sensor Networks, IEEE Transactions On Computers, Vol. 61, No. 2,
February 2012
[21] S. Nithyakalyani, ―Data Relay Clustering Algorithm for Wireless
Sensor Networks: A Data Mining Approach, Journal of Computer Science 8
(8): 1281-1284, 2012
[22] Khushboo Sharma, -Nearest Neighbour Classification for Wireless
Sensor Network Data, International Journal of Computer Trends and
Technology, volume2, Issue-2, 2011
[23] AshuGupta and R.Annie Uthra, Cluster Based Approximate Data
Collection in WirelessSensor Network, IJECSE, Volume2, Number 2
[24] Jyoti Rajput, A Survey on Secure Data Aggregation in Wireless Sensor
Network, IJARCSSE, Volume 4, Issue 5, May 2014
[25] LaxmiChoudhary, ―Challenges for Data Mining, IJREAS, Volume 2,
Issue 2, February 2012
[26] Mukesh Kumar Jha, T.P Sharma ―Secure Data aggregation in Wireless
Sensor Network: A Survey, International Journal of Engineering Science and
Technology, ISSN: 0975-5462, Vol. 3 No.3, March-2011
[27] P. D. Patel, P.B. Lapsiwala, R.V. Kshirsagar ―Data Aggregation in
Wireless Sensor Network, International Journal of Management, IT and
Engineering, vol. 2, Issue 7 July- 2012
AUTHORS PROFILE
Ms.S.Shylaja is an Assistant
Professor & Research Scholar in
Rathinam College of Arts & Science,
at Affiliated to Bharathiar University.
Coimbatore, India. She has more than
two years of experience. Her area of
interest is Data Mining. And she is
published two papers in International
journals & presented 8 papers in
National and International Conferences.
Dr.R.Muralidharan is working as
Vice Principal & Head of the
Department of Computer Science in
Rathinam College of Arts and
Science, Coimbatore, India. He got
his Doctoral Degree in Computer
Science, And also he has got sixteen
years of experience in Academics. He has published papers in 7
Journals and 12 Conferences. - See more at:
http://www.rathinamcollege.edu.in/rmuralidharan/#sthash.uvPCanma
.dpuf
