With the development of Web 2.0, a huge amount of user generated data in social media sites is attracting the attentions from different research areas. Social media data has heterogenous data types including link, text and spatial-temporal information, which poses many interesting and challenging tasks for data mining. Link is the representation of the relationships in social networking sites. Text data includes user profiles, status updates, posted articles, social tags, etc. Mobile applications make spatial-temporal information widely available in social media. The objective of my thesis is to advance the data mining techniques in the social media setting. Specifically I will mine useful knowledge from social media by taking advantage of the heterogenous information including link, text and spatial-temporal data. First, I propose a link recommendation framework to enhance the link structure inside social media. Second, I use the text and spatial information to mine geographical topics from social media. Third, I utilize the text and temporal information to discover periodic topics from social media. Fourth, I take advantage of both link and text information to detect community-based topics by incorporating community discovery into topic modeling. Last, I aggregate the spatial-temporal information from geo-tagged social media and mine interesting trajectories. All of my studies integrate the link, text and spatial-temporal data from different perspectives, which provide advanced principles and novel methodologies for data mining in social media.
【 预 览 】
附件列表
Files
Size
Format
View
Exploring link, text and spatial-temporal data in social media