|Author:||Wong, Man Yung|
|Title:||Scalable topic-based affinity propagation implementation|
|Subject:||Online social networks.|
Digital communications -- Social aspects.
Internet -- Social aspects.
Hong Kong Polytechnic University -- Dissertations
|Department:||Department of Computing|
|Pages:||78 leaves : illustrations ; 30 cm|
|Abstract:||Social networks are important for daily live. People exchange thoughts, ideas, experience, emotions, everything through social networks. Social networks linked up people who have the same thought, the same family, the same school, the same company, etc. Before technologies were not well-developed and people can only contact each other through writing letter, talking on phone and even talk to each other face-by-face. It obstructs people to keep up the connection to each other inside social networks. So messages are possible be lost due to a broken connection. However, people are still able to maintain network connection by putting extra effort. In this century, online social network changes the way in how we interact to others. People are able to spread messages on social network platforms like Facebook, Twitter, Youtube and Weibo etc. People could send a message in the new channel, the Internet, the online social network platforms. Benefit from the era of technologies, there are many instant messenger applications that are available as well. People are connected. Beside these efficient and effective channels, the mobile technology made big changes to human life. It allows people to send out the message, share the message, in anytime, anywhere. Also, we are huge social networks which may have more than million people. To analysis those huge social networks, we are able to monitor human behavior and predict the trends.|
In this paper, we are following Topical Affinity Propagation (TAP) which proposed by Jie Tang. Topical Affinity Propagation is a robust topic based social network analysis method. It can produce a rating per person to different topics in a social network. Beside the Topical Affinity Propagation approach we also investigate a new approach to solve the problem to analyze a large-scale graph (a large social network) in an efficient way by running TAP on an open source engine called Spark. Spark is an open source cluster system that is able to improve the data analysis processes by utilizing the resources (e.g. CPU power, Memory, Hard disk storage, etc.) from different computer in the cluster. We setup the Spark cluster and carry out the TAP algorithm in different size of the social network graph in the experiments. Because of the resources limitation, we do the experiments with three computers. We performed three experiments in the in this research. In our experiment, we start with a small data set in the first experiment, a large data set in the second experiment and a huge data set in the third experiment. Our approach is able to overcome the hardware limitation of large data set analysis. For instance, the memory issues in process a large data set, the limited processing power. However, there are much room that we can improve our approach. We can improve the iteration in the code level, have a better approach to partition the data and have a better algorithm for the node score calculation.
|Rights:||All rights reserved|
Files in This Item:
|b27576747.pdf||For All Users (off-campus access for PolyU Staff & Students only)||1.82 MB||Adobe PDF||View/Open|
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: