A|T|C|T
Application and Theory of Computer Technology
Article
Neural Learning Technique Inculcation for
Commended Social Network Rating (CSNR) System
K.T.Senthil Kumar 1 and Dr.R.Ponnusamy 2
1
2
Part-Time Research Scholar in Computer Science, R&D Department, Bharathiar University, Coimbatore,
India. [email protected]
Department of Computer science and Engineering, Sri Lakshmi Ammal Engineering College, Chennai,
India. [email protected]
Received: 03-January-2017; Accepted: 18-January-2017; Published: 22-January-2017
Abstract: Broadcasting the online social networks, the main aim over proposed suggestions of social
websites became desperate for victory of online applications signifying role in accessing user
preference. Though user preference executed data deficiency and cold start problem immerge
together causing major defect against recommender system. Existing techniques used matrix
factorization method to avoid data lacking yet the process used many deviated methods to stop it
completely affecting the trustee relationship. In our proposed system the intention is to explore
trustee information in commended system and the choice made significantly persuading social
content. We propose a neural learning technique for providing commendation in Commended
Social Network Rating (CSNR) the promotion also incorporates trustee and distrust relation and its
learning. The main aim is to incorporate the neural learning technique that classifies the positive
and negative feedbacks from the trustee and according to their prediction the social network rating
will be done. The character prediction and analysis of trustee helps in finding their recommendation
to be taken in apt way or they might be categorized into always positive or negative segregation.
Keywords: Commended Social Network Rating (CSNR), recommended system approaches, trustee
and distrust relation, neural learning
1. Introduction
Nowadays there are many demand for online oriented services in social networking. Each and
every user have specific needs and interest in various fields it may be share market, economics, sports,
politics, movies, music, medical, software field etc. Web provides enormous amount of information,
making it more challenging for the user to find relevant solution the customer interested in.
proposal\recommender system devote to issue user with suggestion of products they like,
considering their previous rating, cookies or involvement in particular thing. Taking this as a note
many online applications (Amazon, flip kart, iTunes and so on) seek their success by evaluating and
fulfilling the user recommendation. The social network are classified as:

Explicit social n/w: ESN is a correlation provided by the user who gets connected explicitly
defining network with user based on their post/interest. Some of explicit network are Facebook,
Myspace and twitter.

Implicit social n/w: ISN is a correlation implied from user action defined by his/her own interest
the connection is not explicitly connected, but purely related to the individuals interest
exemplified by online behavior. Some of the implicit networks are co-worker, email network
and color.
To pursue user needs the individual must have personalized interest in online updates. Though
user have needs for search they are not ready to spend time to explore their personal information
27 | P a g e
Appl. Theory Comput. Technol. 2017, 2, 1; doi: 10.22496/atct20170103123
A|T|C|T
Application and Theory of Computer Technology
longings. The goal of recommended system is to identify the information relevant for user by
originating neural learning technique and to share posts relating the user likes/dislikes.
Social network evolution affected by certain effects includes self-interest, social/resource
exchange, balance and proximity. Cold start problem is frequent problem in commended system,
they are recent user who have explored only minimum rating making it harder for the system to
collect the recommended needs. CS information are filtered that are likely to concern of user. Usually
there are two approaches implemented in commended systems They are content based approach, the
system tries to suggest data corresponding the previous likes and comments of the user. The
suggestions are provided by querying the user needs explicitly or implicitly by surveying the user
behavior.
Collaborative filtering (CF), In CF the process is reversed the commended system identifies user
who share similar product with another active user and enhance items of like-minded user. Owing
to the cold start issue CF may fail if no one in online have rated in past. By calculating over contentoriented commended system, implies filtering based approach that do not use explicit representation
stationed on rating the user. Due to increasing popularity and demand in products data deficiency
and performance drop may occur. This can be sorted out by increasing the efficiency and
performance of commended system. As result of alleviating this expectation in exploiting social
information emphasizing trust relation amidst users by rating data improving the efficiency of
commended system.
Matrix factorization model used previously for rating commended system in social rating
network (SRN) may incorporate trustee and distrust relations targeting to provide good quality
suggestions that ease the deficiency and cold start issue. Though the factorization technique provides
efficiency time delay takes place for understanding the user needs making it difficult for the system
to counterpart the trust relation and distrust relation. To regulate optimization and performance
efficiency regularization based social recommendation implemented for further usage. While
implementing previous method the optimization issue is tough to recover.
In accordance with Gediminas et al (2005) the recommender systems of present recommendation
methods classifies content based approach, collaborative approach and hybrid classifications
extending the broader range of classifications. The context based information processing the
recommendation for support the ratings of flexible and intrusions.
Bobadilla et al (2015) represented the web development in parallel social information system
that collaborate method filtering the algorithms. The evolution process classifies system for
identifying the implementation for developing the forecasting of future methodologies where they
require implicit information over filtering of recommended systems.
In our proposed paper we implicate various sections to provide betterment at present. The
following commended system we represent factors influencing social rating network, section two
deals with trustee and distrust recommendation system. Section three composes recommendation
neural learning and section five accompanies approach for recommendation in social networks along
with classifications of recommendations is explained. Section six includes the experimental and result
analysis of the recommended system. As the last section we made a conclusion over commendation
in social networking system. Thus our approach towards the rating of online system over
commended prediction with positive and negative analysis. Apart from rating the trusted
recommender will be character analyzed thus to assume their predictions value.
2. Factors influencing social network rating
Rating prediction is held predicting the ratings of the target user for targeted items/product. The
rating method is validated considering the top-N items recommendation predicting the highest rated
products amidst the product that is not yet rated by the fixed user. To execute the system, a rating
28 | P a g e
Appl. Theory Comput. Technol. 2017, 2, 1; doi: 10.22496/atct20170103123
A|T|C|T
Application and Theory of Computer Technology
matrix is stimulated for gaining the perspective results of user needs recommending certain items or
products.
Target user
products
Cricket
Football
Hockey
Tennis
kabadi
A
2
1
4
2
5
B
5
1
?
4
1
C
4
5
2
?
?
D
2
2
4
2
?
Similar user
TABLE 1. Social n/w rating matrix.
The commended system (CS) predicts the user interest by analyzing their characters or
personalization of the user. For instance, in above table as described the social networking rating
formulates a clear view initializing user’s preference that user A aspects rating for hockey viewing
simultaneously user D have similar interests as same as A do. Every user have specific emotions the
users interest depends on their character. hence, accordingly the recommendation can be made in
this the system attempts to send alternative information regarding sports that to particularly
collection of information about hockey SNR estimates user needs by their behavior depending on
like, dislikes, share, posts, and comments or even by views of the particular user.
2.1 Validating user profile
Validation process is nothing but predicting what type of recommendations the user may
expect? The views and expectations may vary from user to user, it is not necessary that every point
of view of user must be the same. To capture the user’s mind the commended system should go
through the profile of user uploaded in any social networking. Regarding the comments and
suggestion the system comes to an end about the preference the individual likes.
2.2 Deviating positive and negative recommendation
As discussed earlier user have different thoughts and behavior. The task of recommended
system is not only to view and estimate user’s profile. CS should have the capacity of deviating views
of user. Consider user A as an aspect, user A having an account in any of the social networking site
may be Facebook, twitter or any other common site would share, like or comment on any information
he/she interested in. While CS progressing the user profile the system may get hints depending the
user’s pursuit but CS must validate the pursuit in depth whether the comment or share the user pass
is in positive view or in negative view. If user A shares information about any particular media actor
then CS must validate the share is in negative way or positive way, it may be trolls of the actor or it
may be regarding the achievements of the actor. On basis of achievements related then recommender
must collect good things about the actor and share to the user.
29 | P a g e
Appl. Theory Comput. Technol. 2017, 2, 1; doi: 10.22496/atct20170103123
A|T|C|T
Application and Theory of Computer Technology
Figure.1 Deviation in user interest through recommendation system
Similarly by scanning one user profile the system could be able to collect details of user friend’s
likes and friends of friend’s likes and dislikes by posting several other information inducing products.
3. Recommendation set and approaches in social network
Recommendation always provides appraisal reports according to the user review that propose
recommendation where the recommendation data set helps in accommodating social network. The
two data sets for commended system in social network are:
(1) EPINIONS: Epinions are defended as common consumer reviewing sites. As acquisition the
visitor using Epinions can read old and new reviews elevating various products. This site mainly
acquired to help visitors choose their purchase. The product which gets more positive reviews
will be the decision for purchase. Epinions are simply denoted as online product review that is
mentioned as explicit perception of trust. The chances of hike in rating may get decided as users
express their trust on reviewers mostly. Whereas in part of cold start only 50% newcomers occur
and that too rating is less than 5.
(2) FLIXTER: flixter is a social networking movie site enhances new movie release, learning about
movies and there is also a chance of meeting other user with same taste. Through this an
interconnecting relationship/friendship is formed. In flixster 85% user have no ratings and 50%
are cold start raters with only 5 ratings.
3.1 Approaches in recommended system
Some basic approaches are followed in recommendation social network classified as memory
oriented and model oriented approach.
(1) Memory oriented recommendation often explore social network for raters and stores the social
rating details. To compute information prediction it aggregate the ratings and provide zero
learning phase. Optimize finding raters in neighborhood of active user who use different method
to calculate the trusted relation. But enhance slow prediction the first generating recommenders
in social network used memory oriented approach.
(2) The next approach in commended system is model based approach exploits appraisal by learning
a new model stores model parameters with the highly recommended sites. Model based
approach is fast in prediction and prefers substantial time for immense learning. In model
orientation most methods are based on matrix factorization.
30 | P a g e
Appl. Theory Comput. Technol. 2017, 2, 1; doi: 10.22496/atct20170103123
A|T|C|T
Application and Theory of Computer Technology
4. Neural learning trustee relation and distrustee relation
In recommended system the suggestion the system gained about the user may be trustee relation
or distrust relation. The system may not blindly believe the mid suggestion it takes several process
to gain the end result. Trustee network permits user to screen which user have declared trust relation
in another user. Trustee network systematically document their relationship (trust). Trustee user
control other user the interconnected user do not have any social relationship. Mainly the trust in
user is dependent on reviews, tags and articles designed by the user itself. The effects of SRN are
measured for influence purpose.
4.1 Influence in social network
Social power is influenced by predicting the ratings of friends. As friends likely have similar
rating than others. In correlation influence the influence is done by predicting ratings of actors with
likely rating if certain ratings are similar further ratings also be similar. The ratings in selection
reactors relate to reactors with identical ratings. The last effect is transitivity that the actors related to
friends of their friends they can also relate to indirect neighbor.
The major benefit of recommendation influence is they have capacity to interact with cold start
user, while they are connected to the social network. They show robustness in finding fraudulent
activities when a profile is attacked. In case of challenges trustee network provide probability
detecting the rater of intended item coverable at small distance. It becomes more challenging when
ratings are done at large network distance creating noise. The data updated in social network is high
sensitive implemented with privacy concern. Hence the information gathered should be more
confidential. A social influence plays main role in time activation depending the user time. The
trustworthy relation maintains the system using trustee relationship imports certain parameters for
validation process listed below.

TIDAL TRUST: tidal trust implements altered breath first search in network. Consider all raters
at minimum distance from intended user. The trust depends on connecting paths. Tidal trust
prefer raters efficiency, precision, low recall all at minimal distance.

MOLE TRUST: mole trust is identical to tidal trust. This considers rater to a maximum depth.
The process is held by tuning trade-off between precision and recall.

TRUST WALKER: Trust walker based on random walk method that induces in exploring the
network. Trust walker use only known and near neighbors who have rated the intended item
instead of using far neighbor. In trust walker we denote as combination of item based
recommendation and trust based recommendation. Several random walks are performed on
network. Each walker returns a rating of similar item. Here, prediction is equal to aggregate
returned ratings.
METHOD
Item based
User based
Mole trust
Tidal trust
Walker
RMSE
1.5
1.4
1.4
1.2
1.2
%
21.2
10.3
55.3
56.9
70.7
CSU
F-measure
0.3
0.2
0.5
0.6
0.7
RMSE
1.2
1.2
1.2
1.1
1.0
All users
%
F-measure
68.9
0.6
67.4
0.6
81
0.7
83
0,7
94
0.8
Table 2: RMSE results on Epinions and flixter
31 | P a g e
Appl. Theory Comput. Technol. 2017, 2, 1; doi: 10.22496/atct20170103123
A|T|C|T
Application and Theory of Computer Technology
The above techniques are formulated by neural learning. Neural learning is an automatic
learning done by calculating the user interest and recommended system prediction.
Figure.2 Commended social Network Rating Architecture
The above figure depicts the entire commendation system architecture where the CSNR system
recommends the social sites. The trustee and trusted members references and suggestions will be
taken and consider for rating them accordingly. Some will be recommended while some sites will be
evaluated as less and ratings will be given negative. Our proposed commended rating system will
not accept the rating recommendations from trustee intact. It will cross check the recommender’s
personal profile including their type of conversations, replies, their complete character and emotional
behavior will be noted and accordingly their suggestions will be validated.
The positive or negative feedback rating will be updated after analyzing it in all the possible
ways. Then the results will be appended in a table for future orientation and reference separately.
The trustee feedback will be in a table that refers whether the suggestions from a particular person
or a group are valid or not. This will be finalized according to the review result of emotional and
sentimental analysis done on that particular personnel profile or that group conversation will be
taken for analysis. Their views and perspectives on other issues will be analyzed and accordingly
their rating will be done.
As a main part in neural learning various social websites CSNR will be appended in a dataset
for suggesting them to various other sites. The recommendations will be classified as trendy
recommendations, seasonal recommendations etc. for example the current trends and issues or any
global issues like any war between countries, any personalities personal life problems or their death,
economical issues etc. all these are known as trendy recommendations. This kind of
recommendations suggests recent and current fashions which will be changing with upcoming and
latest news. Seasonal recommendations are like president elections or local elections in countries,
world beauty contest, Olympics, T20-20 match, world cup match etc. as they come regularly after
certain period of time. In the commended rating of social network helps in secure and cynical
activities in web oriented activities.
32 | P a g e
Appl. Theory Comput. Technol. 2017, 2, 1; doi: 10.22496/atct20170103123
A|T|C|T
Application and Theory of Computer Technology
5. Result analysis
The Commended system rating the social network websites will be more accurate than any other
rating system as it recommends not only on the basis of suggestions and recommendations but also
by accomplishing complete emotional and sentimental analysis on their personal profiles. Each
trustee will be validated by performing analysis over their comments, follows, likes and their
postings in newsfeeds. If they use more positive words and positive comments and suggestions then
they are classified as good even though provided negative feedbacks for wrong things then they are
rated as high trustee.
In case of more negative people always criticizing about something and somebody and always
uses negative comments and negative kind of feedbacks and suggestions then they will be rated less
as their nature makes them give more negative suggestions than positive. Thus their references will
be given less importance. The datasets regarding record of number of social networking websites and
their followers will be listed as follows.
Figure.3 Social networking sites list for users and types
The social sites will be analyzed with all the sites and they are classified as trendy suggestions
and seasonal suggestions while acquiring in common. When searches in particular the entire list of
suggestions will be displayed as in common. The most trending and current news will be updated
33 | P a g e
Appl. Theory Comput. Technol. 2017, 2, 1; doi: 10.22496/atct20170103123
A|T|C|T
Application and Theory of Computer Technology
and suggested at first then if any seasonal occasions or games or events undergoing then that will be
exhibited. According to the suggestions those sites will be mostly viewed or used by the internet
users frequently. Most of the user suggestions helps in safe surfing and provides latest and complete
updates.
5. Conclusion
In our proposed paper we recommend a Commended Social Network Rating system that
facilitates rating and suggesting trending social sites for current news. The seasonal and trending
newsfeed data will be updated and recommended to the frequent internet users through
commendations given by this rating system in social networking websites. Through complete
analysis of emotions and sentiments from the personal profile classification and according to the
trusted recommenders references internet users can browse securely. It also assists them with new
approach towards websites by providing them safe, secure and frequent surfing through latest
websites. It makes them more updated not only with trendy and seasonal revision of information but
also keep them away from phishing websites by providing negative suggestions on the unsecured
websites. As internet and social websites usage and creations growing rapidly in future we need to
enhance the system with defense over any attacks or any forthcoming attacks.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
Gediminas Adomavicius and Alexander Tuzhilin. Toward the next generation of recommender systems:
A survey of the state-of-the-art and possible extensions. IEEE Transactions on Knowledge and Data
Engineering, 17(6):734–749, 2005.
Deepak Agarwal and Bee-Chung Chen. flda: matrix factorization through latent dirichlet allocation. In
Proceedings of the third ACM international conference on Web search and data mining, pages 91–100. ACM, 2010.
Paolo Avesani, Paolo Massa, and Roberto Tiella. A trust- nhanced recommender system application:
Moleskiing. In Proceedings of the 2005 ACM symposium on Applied computing, pages 1589– 1593, 2005.
Giacomo Bachi, Michele Coscia, Anna Monreale, and Fosca Giannotti. Classifying trust/distrust
relationships in online social networks. In Privacy, Security, Risk and Trust (PASSAT), 2012 International
Conference on and 2012 International Confernece on Social Computing (SocialCom), pages 552–557. IEEE, 2012.
Jesús Bobadilla, Fernando Ortega, Antonio Hernando, and Abraham Gutiérrez. Recommender systems
survey. Knowledge-Based Systems, 46:109–132, 2013.
John S Breese, David Heckerman, and Carl Kadie. Empirical analysis of predictive algorithms for
collaborative filtering. In Proceedings of the Fourteenth conference on Uncertainty in artificial intelligence, pages
43–52.Morgan Kaufmann Publishers Inc., 1998.28
Moira Burke and Robert Kraut. Mopping up: modeling wikipedia promotion decisions. In Proceedings of
the 2008 ACM conference on Computer supported cooperative work, pages 27–36. ACM, 2008.
Dorwin Cartwright and Frank Harary. Structural balance: a generalization of heider’s theory. Psychological
review, 63(5):277, 1956.
Gang Chen, FeiWang, and Changshui Zhang. Collaborative filtering using orthogonal nonnegative matrix
tri-factorization. Information Processing &Management, 45(3):368–379, 2009.
Andrew Cotter, Ohad Shamir, Nati Srebro, and Karthik Sridharan. Better mini-batch algorithms via
accelerated gradient methods. In NIPS, volume 24, pages 1647–1655, 2011.
David Crandall, Dan Cosley, Daniel Huttenlocher, Jon Kleinberg, and Siddharth Suri. Feedback effects
between similarity and social influence in online communities. In Proceedings of the 14th ACM SIGKDD
international conference on Knowledge discovery and data mining, pages 160–168. ACM, 2008.
Sanjoy Dasgupta, Michael L Littman, and David McAllester. Pac generalization bounds for cotraining.
Advances in neural information processing systems, 1:375–382, 2002.
Mukund Deshpande and George Karypis. Item-based top-n recommendation algorithms. ACM
Transactions on Information Systems (TOIS), 22(1):143–177, 2004.
34 | P a g e
Appl. Theory Comput. Technol. 2017, 2, 1; doi: 10.22496/atct20170103123
A|T|C|T
Application and Theory of Computer Technology
14.
34.
Thomas DuBois, Jennifer Golbeck, and Aravind Srinivasan. Predicting trust and distrust in social networks.
In Privacy, security, risk and trust (passat), 2011 ieee third international conference on and 2011 ieee third
international conference on social computing (socialcom), pages 418–424. IEEE, 2011.
Rana Forsati, Hanieh Mohammadi Doustdar, Mehrnoush Shamsfard, Andisheh Keikha, and Mohammad
Reza Meybodi. A fuzzy co-clustering approach for hybrid recommender systems. International Journal of
Hybrid Intelligent Systems, 10(2):71–81, 2013.
Rana Forsati andMohammad RezaMeybodi. Effective page recommendation algorithms based on
distributed learning automata and weighted association rules. Expert Systems with Applications, 37(2):1316–
1330, 2010.
Jennifer Golbeck. Computing and applying trust in web-based social networks. PhD thesis, 2005.
Jennifer Golbeck. Generating predictive movie recommendations from trust in social networks. Springer, 2006.
Jennifer Golbeck and JamesHendler. Filmtrust: Movie recommendations using trust inweb-based social
networks. In Proceedings of the IEEE Consumer communications and networking conference, volume 96. Citeseer,
2006.
Nathaniel Good, J Ben Schafer, Joseph A Konstan, Al Borchers, Badrul Sarwar, Jon Herlocker, and John
Riedl. Combining collaborative filtering with personal agents for better recommendations. In AAAI/IAAI,
pages 439–446, 1999.
Quanquan Gu, Jie Zhou, and Chris Ding. Collaborative filtering: Weighted nonnegative matrix
factorization incorporating user and itemgraphs. In SIAM SDM, pages 199–210, 2010.
R. Guha, Ravi Kumar, Prabhakar Raghavan, and Andrew Tomkins. Propagation of trust and distrust. In
Proceedings of the 13th International Conference on WorldWideWeb, pages 403–412. ACM, 2004.
Guibing Guo, Jie Zhang, Daniel Thalmann, Anirban Basu, and Neil Yorke-Smith. From ratings to trust: an
empirical study of implicit trust in recommender systems. In SAC, 2014.
Jonathan L Herlocker, Joseph A Konstan, Al Borchers, and John Riedl. An algorithmic framework for
performing collaborative filtering. In Proceedings of the 22nd annual international ACM SIGIR conference on
Research and development in information retrieval, pages 230–237. ACM, 1999.29
Jonathan L Herlocker, Joseph A Konstan, Loren G Terveen, and John T Riedl. Evaluating collaborative
filtering recommender systems. ACM Transactions on Information Systems (TOIS), 22(1):5–53, 2004.
Thomas Hofmann. Latent semantic models for collaborative filtering. ACMTransactions on Information
Systems (TOIS), 22(1):89–115, 2004.
Mohsen Jamali and Martin Ester. Trustwalker: a random walk model for combining trust-based and itembased recommendation. In Proceedings of the 15th ACM SIGKDD international conference
on Knowledge discovery and data mining, pages 397–406. ACM, 2009.
Mohsen Jamali and Martin Ester. A matrix factorization technique with trust propagation for
recommendation in social networks. In Proceedings of the fourth ACM conference on Recommender systems,
pages 135–142. ACM, 2010.
Mohsen Jamali and Martin Ester. A transitivity aware matrix factorization model for recommendation in
social networks. In Proceedings of the Twenty-Second international joint conference on
Artificial Intelligence-Volume Volume Three, pages 2644–2649. AAAI Press, 2011.
Arnd Kohrs and Bernard Merialdo. Clustering for collaborative filtering applications. In In Computational
Intelligence forModelling, Control & Automation. IOS. Citeseer, 1999.
Ioannis Konstas, Vassilios Stathopoulos, and JoemonMJose. On social networks and collaborative
recommendation. In Proceedings of the 32nd international ACM SIGIR conference on Research and
development in information retrieval, pages 195–202. ACM, 2009.
©
2016
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
35 | P a g e
by
the
authors. Submitted for possible open access publication under the
terms and conditions of the Creative Commons Attribution (CC-BY) license
(http://creativecommons.org/licenses/by/4.0/).
Appl. Theory Comput. Technol. 2017, 2, 1; doi: 10.22496/atct20170103123