Date of Award

Spring 1-1-2016

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

Shivakant Mishra

Second Advisor

Richard Han

Third Advisor

Qin Lv

Fourth Advisor

Tom Yeh

Fifth Advisor

Ahmad Rahmati

Abstract

This thesis investigates new approaches to predict mood and emotional states of individuals and groups of individuals from their online activities in social networks and smartphone usage. In particular, this thesis develops an understanding of the relationship between users’ emotions, their interactions with other people and their activities (personal concerns) and usage in both social networks and smartphones. It demonstrates the feasibility of using attributes from online social networks and smartphones to predict users’ emotions. Mood and emotions of users have a strong impact on their behavior, actions as well as their interactions with other people. They are an important contextual feature for building context-aware pervasive applications. The thesis analyzes two sets of datasets. The first dataset is a large volume of postings from the Twitter social network comprised of profiles and postings of about three million users. The second dataset is collected via a smartphone application used by more than 25 users over a period of more than two months. This application is specifically developed to gather smartphone activities data of the users, including their current emotions, location, activities and applications they are running. Based on the analysis of these datasets, the thesis develops highly accurate classifiers to predict user emotions from their online activities in social networks and smartphone usage.

Share

COinS