Case Study: Virginia Tech forecasts civil unrest with Basis Technology's Rosette

Forecasting the Future The EMBERS Predictive Analytics Success Story

Virginia Tech launched the EMBERS project to forecast civil unrest, elections, disease outbreaks, and other disruptive events in Latin America using open source big data. The challenge was to sift through massive volumes of unstructured text from tweets, news, blogs, and other public sources fast enough to “beat the news,” while incorporating local, country-specific context. Basis Technology’s Rosette text analytics platform was used to enrich and structure this data for downstream forecasting models.

Basis Technology helped Virginia Tech configure Rosette to identify languages, entities, dates, times, locations, and other signals from unstructured text, enabling EMBERS’ ensemble of predictive algorithms to generate alerts. The result was a fully automatic system running 24x7 that digested nearly 20GB of data per day and produced roughly 40–50 warnings daily. By March 2014, EMBERS had delivered more than 10,000 warnings, forecasted major protests in Brazil and Venezuela, and exceeded its two-year goals in three of five evaluation criteria.

Virginia Tech

Naren Ramakrishnan

Director of the Discovery Analytics Center