Data and machine learning in population and public health
Rumi Chunara (NYU)
Tuesday November 12, 12-1pm
McIntyre Medical Building, Room 1034
Abstract: New unstructured Internet and mobile-sourced data from satellite imagery to precisely-geolocated text and images from mobile phones can provide high-resolution spatial and temporal views into daily life factors and prevention efforts important in public health. The low-frequency, low-resolution data commonly used (e.g. surveys, government reports) only provide low-precision behavior and outcome measures on the specific populations sampled. This limits how effectively we can spend public health budgets and target intervention efforts. While it is clear that new data sources have the potential to transform public health efforts, there are critical computational and statistical challenges to making the data useful in aggregate, including non-continuous data shared whenever a person likes, noisy/unstructured observations that may not spatially representative of the underlying (latent) process, and data generated by ad-hoc, non-representative groups. In this talk I will elucidate such challenges and work towards addressing them relevant to public health priorities. Examples will span local and global efforts both in communicable and non-communicable disease.