Masters Thesis Defense “Seasonality in Dynamic Stochastic Blockmodels” By Jace Robinson

Committee:  Drs. Derek Doran, Advisor, Tanvi Banerjee, and Fred Garber (Electrical Engineering)

ABSTRACT:

Sociotechnological and geospatial processes exhibit time varying structure that make insight discovery challenging. This thesis proposes a new statistical model for such systems, modeled as dynamic networks, to address this challenge. It assumes that vertices fall into one of k types and that the probability of edge formation at a particular time depends on the types of the incident nodes and the current time. The time dependencies are driven by unique seasonal processes, which many systems exhibit (e.g., predictable spikes in geospatial or web traffic each day). The thesis defines the model as a generative process and an inference procedure to recover the `normal' seasonal processes from data when they are unknown.
Evaluation is completed on several synthetic and real datasets. The synthetic experiments demonstrate the superior fidelity of this model on seasonal datasets, while being able to remain equally accurate for non-seasonal data. The model is up to twice as accurate at predicting future edge density over competing models on New York City Taxi trips, United States airline flights, and email communication within the Enron company. A software tool named GEONET is developed for anomaly detection and exploration by a human analyst on geospatial data and is utilized on the New York City data.