Skip to content. | Skip to navigation

Sections
Personal tools
You are here: Home Theses Prediction of Community Behavior in News Social Media using Deep Learning

Contact

Prof. Dr. S. Decker
RWTH Aachen
Informatik 5
Ahornstr. 55
D-52056 Aachen
Tel +49/241/8021501
Fax +49/241/8022321

How to find us

Annual Reports

Disclaimer

Webmaster

 

 

Prediction of Community Behavior in News Social Media using Deep Learning

Thesis type
  • Bachelor
Student Svetlana Pavlitskaya
Status Finished
Proposal on 12. May 2016 14:30
Proposal room Seminarraum I5
Add proposal to calendar vCal
iCal
Presentation on 23. Sep 2016 14:00
Presentation room Seminarraum I5
Add presentation to calendar vCal
iCal
Supervisor(s)
Advisor(s)
Recent studies have shown that discussions in social media can become a place for aggressive behavior and attacks especially when the discussions are devoted to politics. Comments are usually uncivil and can lead to attitude polarization. Analyzing user or community reaction on news can help to avoid such situations.

The main research goal of the thesis is the prediction of community reaction on news in social media.

This work aims at exploring how reaction of communities on certain news can be modeled and predicted with deep learning methods. For this, a representative collection of data will be retrieved from public news pages in Facebook. The dataset will be analyzed in order to detect communities and their evolution, retrieve comment sentiments and determine concepts of news. This information will be then used to create a feature set for each time interval when a community exist. The feature set will show the reaction of communities (positive, negative) on particular news (concepts extracted from news posts) as well as include other information such as a number of users, number of left and joined users comparing to the previous community state, etc. Such a time series set will be used to train and test a deep neural network (recurrent neural network) in Tensorflow using 10-fold validation approach.
Document Actions