Feature Selection with Linked Data in Social Media

Feature selection is widely used in preparing high-dimensional data for effective data mining. Increasingly popular social media data presents new challenges to feature selection. Social media data consists of (1) traditional high-dimensional, attribute-value data such as posts, tweets, comments, and images, and (2) linked data that describes the relationships between social media users as well as who post the posts, etc. The nature of social media also determines that its data is massive, noisy, and incomplete, which exacerbates the already challenging problem of feature selection. In this paper, we illustrate the differences between attribute-value data and social media data, investigate if linked data can be exploited in a new feature selection framework by taking advantage of social science theories, extensively evaluate the effects of user-user and user-post relationships manifested in linked data on feature selection, and discuss some research issues for future work.