The Power of a Machine Learning Combined Data Approach: Open Source Data and Population Characteristic Data

Fraym proposes to combine our 1x1km human geography data with publicly available event data in our machine learning models to inform Air Force threat prediction approaches. For this effort we seek to prove out the benefit of using population characteristic data and publicly available event data to model for emerging crises. We will employ our standard best practices in machine learning around data and threat prediction in geographies of high interest to the Air Force and subordinate organizations. The final output will be a scalable algorithm and data model that could translate to other risk-based scenarios with a phase II goal of replication in a new country without current data coverage. Fraym provides geospatially enabled localized data on populations across 69 data poor countries and growing. We leverage professionally enumerated survey data combined with over 40 remotely sensed geospatial covariates in our machine learning models creating predicted surfaces of more than 60 human indicators. By rapidly processing publicly available data combined with our standard data layers, we have successfully created risked based models for issues like risk for terrorism, social unrest, and political violence all of which are essential to monitoring points of interest, mission planning, and overall situational awareness.