Dv31821825

203 views

Published on

IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
203
On SlideShare
0
From Embeds
0
Number of Embeds
3
Actions
Shares
0
Downloads
3
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Dv31821825

  1. 1. G. Padma, Ravi Kumar Athota, Praveen Kumar Padigela /International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.821-825 Analysis of Social Networks - Study & Emergence of Domain Equivalent User Groups G. Padma, Ravi Kumar Athota, Praveen Kumar PadigelaAbstract – Collective behavior is thesocial network sixties (well before the invention of the internet),data generated by social media like Facebook, whohypothesized the likelihood that any pair ofTwitter, and Flickr presents the opportunities actors on the planet are separatedby at most sixand challenges to studyon a large scale. In this to degrees of separation. While such hypotheses havepredict collective behavior in social network largely remainedconjectures over the last fewdata, collect the relevant users from the social decades, the development of online socialnetworksnetwork, about some individuals, observe the has made it possible to test such hypotheses at leastbehavior ofindividuals in the same network and in an onlinesetting. This is also referred to as thecombined into one group. Asocial-dimension- small worldphenomenon. This phenomenonwasbased approach is effective in addressing the tested in the context of MSN messenger data, and itheterogeneity of connections presented in was shownin [2] that the average path lengthsocialmedia. However, the networks in social between two MSN messenger users is 6.6.This canmedia are normally of colossal size, involving be considered a verification of the widely knownhundreds of thousands of users with different rule of “six degreesof separation” in (generic) socialenvironments. Sparse social dimensions, the networks. Such examples are by no meansunique; aproposedapproach can efficiently works to wide variety of online data is now available whichhandle the networks of several actors has been usedto verify the truth of a host of otherperformance to othernon-scalable methods. conjectures such1 as that of shrinkingdiameters [3] or preferential attachment. In general, theKeywords – Social Networks, Learning Behavior, availability of massiveamounts ofdata in an onlineClassification, Clustering settinghas given a new impetus towards ascientific and statistically robust study of the field of social networks.Growing the complexity of current I INTRODUCTION networks, thetask of ensuring robustness and Marketing is one of main knowledge maintaining quality of servicerequires increasinglyapplication mining where a product is promoted powerful network management systems.This steadyindiscriminately to all potential customers attempts increase in network size and complexityproduces ato select the likely profitable. In general, a social corresponding increase in the volume of data,suchnetwork is defined as a network of interactions or as performance indicators or alarms, to berelationships,where the nodes consist of actors, and processedby management systems. In particular, thethe edges consist of therelationships or interactions area of faultmanagement remains a key problem forbetween these actors. A generalization of the ideaof network operators,as the speed at which faults aresocial networks is that of information networks, in handled has very immediateconsequences forwhich the nodes couldcomprise eitheractors or network performance.entities, and the edges, denote the relationshipsbetweenthem. Clearly, the concept of socialnetworks is not restricted to thespecific case of an SECTION IIinternet-based social network such as Facebook; the 2. Related Work:Scalable learning and collectiveproblemof social networking has been studied often behavior refers to the behaviors of individuals in ain the field of sociology interms of generic social networking domain but it is not simply theinteractions between any groups of actors. Such aggregation of individual behaviors connectedinteractionsmay be in any conventional or non- environment to be interdependent influenced by theconventional form, whether they be face- behavior of friends. For example in the domain oftofaceinteractions, telecommunication interactions, marketing friends buy something there is a betteremail interactions or postalmail interactions.The than average chance that we will buy it too, this typeconventional studies on social network analysis of behavior is correlation. The person in the networkhave generally not focused on online interactions, more likely to connect to others who share certainand have historically preceded the advent similarities with other person, this phenomenon hasandpopularity of computers or the internet. A classic been observed not only in the many processes of aexample of this is the studyof Milgram [1] in the physical world but also in online systems. Homophily results in a behavior of correlation 821 | P a g e
  2. 2. G. Padma, Ravi Kumar Athota, Praveen Kumar Padigela /International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.821-825between connected friends and tend to behave thesimilarity between those friends. Social networkmedia enables us to study collective behavior on alarge scale networks, behaviors include a broadrange of actions joining a group connecting toperson clicking on advertisement becominginterested in certain topics like business, educationand entertainment etc. This work attempt to leveragethe behavior correlation presented in a socialnetwork media. Classifications with networked instancesare known as within network classification orspecial case of relational learning the data instancesin a network are not independently identicallydistributed as in conventional data mining. Tocapture the correlation between labels ofneighboring data objects typically a markov Figure 1 shows the proposed Analysisdependency assumption that is the label of one nodedepends on other labels of its neighbors normally a 3.1. Social Network Connectivity:relational classifier is constructed based on the 3.1.1. Social dimension extraction:relational features of labels data and then an The latent social dimensions are extractediterative process is required to determine the class based on network topology to capture the potentiallabels for the unlabeled data. The class label or the affiliations of actors. These extracted socialclass membership is updated for each node while the dimensions represent how each actor is involved inlabels of its neighbors are fixed. A network tends to diverse affiliations. These social dimensions can bepresent heterogeneous relations and the Markov treated as features of actors for subsequentassumption can only capture the local dependency to discriminative learning. Since a network ismodel network connections or class labels based on converted into features, typical classifiers such aslatent groups similar idea is also adopted in different support vector machine and logistic regression canheterogeneous relations in a network by extracting be employed. Social dimensions extracted accordingsocial dimensions to represent the potential to soft clustering, such as modularity maximizationaffiliations of actors in a social network. and probabilistic methods, are dense.SECTION III 3.1.2. Discriminative learning:3. Problem Definition:Communication to discuss The discriminative learning procedure willbusiness education and any activities now days is by determine which social dimension correlates withusing the social networking concept such as Linked the targeted behavior and then assign properIn, Twitter and Facebook. Even some users are weights. A key observation is that actors of the samemissing those activities because all the users are not affiliation tend to connect with each other. Forrelated to such network. For example in instance, it is reasonable to expect people of theFacebookuser only accepts the his/ her friends they same department to interact with each other moremay in different domains and technology to avoid frequently. A key observation is that actors of thethis problem our system generates the other network same affiliation tend to connect with each other. Foronly related to particular activity, the type of activity instance, it is reasonable to expect people of thein mining is association. same department to interact with each other more frequently. Hence, to infer actors’ latent affiliations, we need to find out a group of people who interact with each other more frequently than at random. 3.1.3. Chart Generation for Group/Month: Two data sets reported in are used to examine our proposed model for collective behavior learning. The first data set is acquired from user interest, the second from concerning behavior; we study whether or not a user visits a group of interest. Then generates chart the based on the user visit group in the month. 822 | P a g e
  3. 3. G. Padma, Ravi Kumar Athota, Praveen Kumar Padigela /International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.821-825 3.1.4. Chart Generation for User/Group: Two data sets reported in are used toexamine our proposed model for collective behaviorlearning. The first data set is acquired from userinterest, the second from concerning behavior; westudy whether or not a user visits a group of interest.Then generates chart the based on the user visitgroup in the month.3.2. Algorithm for Learning of CollectiveBehaviorInput: network data, labels of some nodes, numberof social dimensions; Output: labels of unlabeled nodes.1. Convert network into edge-centric view. Figure 2 screen shows different activities of groups2. Perform edge clustering in the Network3. Construct social dimensions based on edgepartition node belongs to one community as long asany of its neighboring edges is in that community.4. Apply regularization to social dimensions.5. Construct classifier based on social dimensions oflabeled nodes.6. Use the classifier to predict labels of unlabeledones based on their social dimensions.Social network mined to discover the tools neededto analyze large complex and frequently changingsocial media data.Social dimensions extractedaccording to soft clustering such as modularity andprobabilistic methods are sparse and dense. Figure 3Before going to participate in the group need to register in the type of network, Here screen SECTION IV shows the successfully registration.4.1. Evaluation on Social dimension: Two datasets reported are used to examine our analysis modelfor collective behavior, first data set is acquiredfrom twitter and second from a popular networkFlickr concerning same behavior. Our analysis aimsto join a group of interest, since the blog catalogdata does interests as the behavior labels. The datasets are publicly available at the first authorshomepage to scalable this network we also include amega scale network Facebook and You tube movethose nodes without connections and select theinterest groups with more than 500 subscribers onour network.Network is based on the linkhttp://www.socialnetworks.mpi-sws.org/data- Figure 4To maintain the social network our analysisimc.html developed Admin, He need to connect the relevant user into one group. 823 | P a g e
  4. 4. G. Padma, Ravi Kumar Athota, Praveen Kumar Padigela /International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.821-825 Our paper motivated the trend towards social networking system such environments social implications must be handled properly. User can interact directly with the relevant activities in human user in the loop numerous concepts, including personalization, expertise involvement,drifting interests, and social dynamics become of paramount importance. Therefore, comparison shows web standard concepts that let people offer their expertise in a service-oriented manner and covered the deployment, discovery and selection of Human- Provided Services. Future work extends at providing more fine-grained monitoring and adaptation strategies, implements the development with novel association mining algorithm.Figure 5 screen shows the several network groups in Reference the collective behavior [1] A. T. Fiore and J. S. Donath, “Homophily in online dating:when do you like someone like4.2. Comparative Study:As existing approaches to yourself?” in CHI ’05: CHI ’05extendedextract social dimensions suffer from scalability, it abstracts on Human factors in computingis imperative to address the scalability issue. systems. New York, NY, USA: ACM, 2005,Connections in social media are not homogeneous. pp. 1371–1374.People can connect to their family, colleagues, [2] L. Tang and H. Liu, “Scalable learning ofcollege classmates, or buddies met online. Some collective behaviorbased on sparse socialrelations are helpful in determining a targeted dimensions,” in CIKM ’09: Proceeding ofthebehavior while others are not. This relation-type 18th ACM conference on Information andinformation, however, is often not readily available knowledge management.New York, NY,in social media. A direct application of collective USA: ACM, 2009, pp. 1107–1116.inference or label propagation would treat [3] P. Singla and M. Richardson, “Yes, there is aconnections in a social network as if they were correlation: - fromsocial networks to personalhomogeneous. A recent framework based on social behavior on the web,” in WWW’08:dimensions is shown to be effective in addressing Proceeding of the 17th internationalthis heterogeneity. The framework suggests a novel conference on World WideWeb. New York,way of network classification: first, capture the NY, USA: ACM, 2008, pp. 655–664.latent affiliations of actors by extracting social [4] M. McPherson, L. Smith-Lovin, and J. M.dimensions based on network connectivity, and Cook, “Birds ofa feather: Homophily innext, apply extant data mining techniques to social networks,” Annual Reviewclassification based on the extracted dimensions. In ofSociology, vol. 27, pp. 415–444, 2001.the initial study, modularity maximization was [5] J. Leskovec, J. Kleinberg, C. Faloutsos.employed to extract social dimensions. The Graphs over time: Densificationlaws,superiority of this framework over other shrinking diameters and possiblerepresentative relational learning methods has been explanations. ACM KDD Conference,2005.verified with social media data in. The original [6] J. Leskovec, E. Horvitz. Planetary-Scaleframework, however, is not scalable to handle Views on a Large Instant-Messagingnetworks of colossal sizes because the extracted Network, WWW Conference, 2008.social dimensions are rather dense. In social media, [7] D. Liben-Nowell and J. Kleinberg. The linka network of millions of actors is very common. prediction problem for socialnetworks. InWith a huge number of actors, extracted dense LinkKDD, 2004.social dimensions cannot even be held in memory, [8] S. Milgram. The Small World Problem,causing a serious computational problem. Psychology Today, 1967.Sparsifying social dimensions can be effective in [9] M.Mcpherson, L.smith-Lovin, and J.M.eliminating the scalability bottleneck. In this work, Cook, “Birds of a feather: Homophily inwe propose an effective edge-centric approach to social network, “Annual review of Sociology,extract sparse social dimensions. We prove that with vol.27, pp.415-444, 2001.our proposed approach, sparsity of social [10] H.W.Lauw,J.C.Shafer, R.Agrawal, anddimensions is guaranteed. A.Ntoulas “Homophile in the digital world: A live Journal case study,”IEEE Internet CONCLUSION V Computing, vol.14 ,pp. 15-23, 2010. 824 | P a g e
  5. 5. G. Padma, Ravi Kumar Athota, Praveen Kumar Padigela /International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.821-825[11] S.A. Macskassy and F.Provost,”Classification in networked data:A tool kit and a univariate case study,”J.Mach.Learn Res., Vol.8,pp. 935-983,2007. G. PadmaAsst Prof B.Tech CSE from Kamala Institute of Technology & Sciences Huzurabad M.Tech SE from Kakatiya Institute of Technology & Sciences Warangal. Currently she is workingat Teegala Krishna Reddy Engineering Collegehaving more than 3 years of experience in Academicand guided UG & PG students. Her research areasinclude Data mining & Data warehousing,Operating System, Network Security. Ravi Kumar Athota Asst Prof M.Tech Information Technology from SRM University Chennai BE from GITAM College of Engineering. Currently he working at Samskruti College of Engineering & Technologyhaving more than 4 years of experience in Academicand guided many UG & PG students. His areas ofspecialization includes Networking, ComputerNetworks, Operating System. Praveen Kumar Padigela Asst Prof M.Tech Computer Science currently he working at Samskruti College of Engineering& Technology having more than 3 years of experience in Academic and guided UG & PG students. His area of specialization is Mobile Computing, researchareas are Software Engineering, Software ProjectManagement, Design Patterns. 825 | P a g e

×