Graduation Year

2019

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Engineering Computer Science

Major Professor

Adriana Iamnitchi, Ph.D.

Committee Member

John Skvoretz, Ph.D.

Committee Member

Kingsley A. Reeves, Ph.D.

Committee Member

Yao Liu, Ph.D.

Committee Member

Paul Rosen, Ph.D.

Committee Member

Giovanni Luca Ciampaglia, Ph.D.

Keywords

data mining, migration, social network analysis, social phenomena, team formation

Abstract

Online communities exhibit dynamic social phenomena that, if understood, can both influence the design of technical platforms and inform theories about general social dynamics. With increasing popularity, online games provide a rich recording of social dynamics that can contribute to understanding human behavior. This dissertation studies two phenomena of social dynamics at large scale using data traces from online games. The first phenomenon is team formation and the second is players mobility between gaming servers.

This dissertation first presents a framework for collecting data from online gaming through crawling. It includes the data sources and the tools used for data collection and processing. We developed a web crawler to perform longitudinal data collection. We discussed primary design considerations and challenges in data collection that we encountered in this effort.

We examined several hypotheses about team formation using a large, longitudinal dataset from a team-based online gaming environment. Specifically, we tested how positive familiarity, homophily, and competence determine team formation in Battlefield-4, a popular team-based game in which players choose one of two competing teams to play on. Our dataset covers over two months of in-game interactions between over 380,000 players. We showed that familiarity is an important factor in team formation, while homophily is not. Competence affects team formation in more nuanced ways: players with similarly high competence team up repeatedly, but large variation in competence discourages repeated interactions. In addition, we formulated the team formation behaviors into a sign prediction problem. We classified interactions in online team-based games into different classes. Then, we modeled two predictionsign prediction scenarios: teams versus squads. We extracted time-based features from these determinants to show that our determinants are effective in predicting signs between gamers, indicating that prior interactions between gamers (familiarity) are more likely to accurately predict signs.

Finally, we presented a data-driven study focused on characterizing and predicting the mobility of players between gaming servers in two popular online games, Team Fortress 2, and Counter Strike: Global Offensive. Understanding these patterns of mobility between gaming servers is important for addressing challenges related to scaling popular online platforms such as server provisioning, traffic redirection in case of server failure, and game promotion. We built predictive models for the growth and the pace of player mobility between gaming servers. We showed that the most influential factors in predicting the pace and growth of migration are related to the number of in-game interactions. Declared friendship relationships in the online social network, on the other hand, do not affect predicting mobility patterns.

Besides providing a large-scale, empirical based understanding of social phenomena, our work can be used as a basis to a variety of application scenarios including peer, team, group, and server recommendations in gaming platforms.

Download

Share

COinS