Job Market Candidates


Information Systems

Jessica Clark

Jessica Clark Personal website
Advisor: Foster Provost
Dissertation   Connecting Human Behaviors and Demographic Characteristics using Massive Individual-Level Data and Predictive Analytics

Abstract   Modern technologies provide massive amounts of data which reveal both individuals' behaviors and their personal characteristics. My dissertation focuses on the use of predictive modeling to improve our understanding of the ways in which behaviors are connected to demographic and social characteristics. Two papers explore the problem of understanding individual and group TV viewership dynamics using data comprising individual-household-level viewership and household members' personal attributes. The first study develops a model that estimates the likelihood of each member of a multi-person household watching TV. Evaluation for this problem is a particular challenge because the data contain no ground-truth labels, and so we also derive a series of related tasks which indicate success on the core problem. The second study illustrates a probabilistic model which relaxes some of the assumptions inherent in prior models and thus has superior performance on a broader set of tasks. The third study is a design science paper that investigates the efficacy of using matrix-factorization-based dimensionality reduction as a technique to improve predictive modeling using social media data.  
Research   My research uses machine learning techniques and the individual-level data afforded by modern information technologies to explore the relationship between people's personal characteristics and their behaviors. I am primarily interested in two approaches: first, developing predictive models to make novel behavioral inferences in the application areas of television advertising, crowdfunding, and finance; second, design science research evaluating the best practices for developing predictive modeling systems with massive fine-grained data sets to learn about individuals’ behaviors and their individual characteristics. My research program has been inspired by my love of pop culture as well as my past work experience in advertising.

Keywords   Data science, business analytics, design science, advertising, natural language processing, television, social media, crowdfunding, dimensionality reduction, diversity

Xuan Ye

Xuan Ye Personal website
Advisor: Sonny Tambe
Dissertation   Managing Non-Wage Incentives for Digital Production
Research   My primary research interests lie at the intersection of IS and economics (particularly labor economics, organizational economics and urban economics). In my dissertation, I develop new, detailed measures of companies’ management practices by collecting novel data and employing state-of-the-art analysis techniques on these data sets. Using these measures, my research papers establish several empirical findings around a central theme. Modern firms often rely heavily on non-pecuniary incentives to motivate technical employees, which weakens the boundary between employees’ leisure and work. I study how the use of this form of compensation affects firms and workers.

Keywords   Incentives in IT Production, Non-Wage Compensation, HR Analytics, HR Management, Open Source Contributions, Geography of IT Production

Operations Management

Yuqian Xu

  Personal website
Advisor: Mike Pinedo
Dissertation   To be updated.
Research   To be updated.

Keywords   To be updated.


Wei Fu

Wei Fu Email:
Advisors: Jeff Simonoff and Patrick Perry
Dissertation   Nonparametric Method in Statistical Learning: Unbiasedness in regression tree, Survival tree and Estimating the number of clusters
Research   My primary research interests are in nonparametric statistical learning methods, both in supervised learning setting and unsupervised learning setting. In specific, I investigated the bias effect in terms of selecting splitting variable in traditional regression tree (CART) and developed an unbiased regression tree for longitudinal data and clustered data. I also proposed two survival tree algorithms that can handle left-truncated and right censored survival data, as well as survival data with time-varying covariates. For unsupervised learning, I developed a new method for estimating the number of clusters based on cross-validation.

Keywords   Statistical machine learning, Clustering, Regression Tree, Survival Analysis