A Digital Platform for Local Foodborne Illness and Outbreak Surveillance
Foodborne illness affects 1 in 4 Americans, annually. However, only a fraction of affected individuals seek medical attention. In this presentation, we will discuss our collaboration with local public health departments to develop a foodborne disease surveillance platform to supplement ongoing surveillance efforts. The platform currently uses digital data from Twitter and Yelp. We developed a machine learning classifier to differentiate between relevant and irrelevant data. The classifier had an accuracy and precision of 85% and 82%, respectively based on an evaluation using 6084 tweets. These performance results are promising, especially given the similarities between the data classes.