We analyze over 500 million Twitter messages from an eight month period andfind that tracking a small number of flu-related keywords allows us to forecastfuture influenza rates with high accuracy, obtaining a 95% correlation withnational health statistics. We then analyze the robustness of this approach tospurious keyword matches, and we propose a document classification component tofilter these misleading messages. We find that this document classifier canreduce error rates by over half in simulated false alarm experiments, thoughmore research is needed to develop methods that are robust in cases ofextremely high noise.
展开▼
