Social Data Analysis by Non-Linear Imbedding
Abstract
Political science datasets contain information of interest to planners seeking to predict international relations. The goal of this project was to use modern data mining techniques to determine whether such data exists and, if so, to characterize it. We developed a new approach for such analysis based on geometric harmonics. At the heart of our approach is the observation that such relationships are inherently nonlinear and that the data are noisy and incomplete. To demonstrate the power and usefulness of our techniques the focus was on United Nations voting data. It was shown that major historical events could be inferred from these data; that other (linear) techniques did not suffice; and that they could be extended to understanding certain aspects of international relations. We conclude that the project was successful in opening up the field of computational international relations.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 20, 2013
- Accession Number
- AD1013125
Entities
People
- Steven W. Zucker
Organizations
- Yale University