Date of Award

Spring 1-1-2016

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

Qin Lv

Second Advisor

Richard Han

Third Advisor

Shivakant Mishra

Fourth Advisor

Sangtae Ha

Fifth Advisor

Xinyu Xing

Abstract

Nowadays, the near-ubiquitous availability of smartphones and the significant improvement of cellular networks make the video chat applications become mainstream for mobile devices. Meanwhile, because of the capability to make friends in the virtual domain, online random video chat services such as Chatroulette and Omegle have become increasingly popular. Given these changes, we expect the mobile random video chat services will also gain the public attention and greatly increase in volume and frequency soon. In this thesis, I focus on analyzing the user behavior and seeking for possible improvements of user experience in such kind of mobile service. I build an Android-based Omegle compliant mobile random video chat application to collect data at scale. Using the collected data, we analyze user behavior patterns from multiple aspects and reveal some concerns regarding user experience in such service. We then conduct an in-depth meaningful user behavior analysis to understand the key characteristics of effectiveness for promoting long video chat sessions. Furthermore, motivated by the negative user experience caused by the existence of obscene content, I propose an accurate and efficient misbehavior classifier. The classifier leverages multimodal sensors and temporal modality in each session to improve accuracy. It also applies a multi-level cascaded classification procedure to quantify the tradeoff between efficiency and accuracy. Finally, I briefly introduce the potential directions which could be further investigated to improve user experience of mobile random video chat services in the future.

Share

COinS