Abstract:
In recent times, Social Networking Sites (SNS) such as Facebook, Twitter, Foursquare and
IMDb have become major platforms of communications for users in the web. These SNS
allow a user to share ideas, thoughts, and opinions with her friends, family and acquaintances.
Every day millions of newsfeeds and tweets are posted in these SNS. The contents of these
newsfeeds and tweets provide a rich platform for the researchers to identify cognitive and
psychological attributes such as personality, values, and preferences of involved users. Several studies have been conducted to identify these attributes from social media usage. These studies predict psychological attributes (i.e., personality and values) independently by analyzing the contents of the social media usage. None of the earlier research investigates how these psychological attributes combinedly influence users’ behavior and decision making process. Our research will take a step forward to investigate whether these psychological attributes derived from social media interactions correlate, change, and influence one another in real life. This dissertation particularly addresses the following problems by analyzing the social media interactions: i) identifying values from multiple interaction features: we develop a technique for computing unified value score of Facebook users from their different interaction features such as statuses, page-likes, and shared-links, ii) predicting the change of values: we build a hybrid time series based machine learning model to capture the change of values from users social media usage, iii) identifying psychological groups: we identify psychological group of users from their interactions in an egocentric social network, and iv) predicting users’ preferences from psychological attributes: we conduct two case studies for predicting users’ eat-out and movie preferences from their psychological attributes inferred from social media usage.
To address the above research problems, we use users’ social media interactions and psychological attributes to build different models. These models take psycholinguistic attributes as independent variables and real life preferences as dependent variables. We conduct experiments with 726 Facebook users, 731 Twitter and Foursquare users, and 330 IMDb users to build and validate our machine learning models. Our models achieve moderate to strong prediction potentials in predicting users’ psychological attributes from their social media usage.
These models also outperform the current baseline approaches and our validation results are
consistent with the real world.
Keywords: Social Networks; Data Analytics; Personality, and Basic Human Values
