Why Google Flu doesn’t accurately predict the flu
- people are changing their search terms (maybe getting better)
- searching the flu isn’t actually indicative of the flu
- having more or less publicity influences searches
- maybe people are differentiating from the cold
- we’re not exactly sure how they built their model or if they are changing
- are they testing and updating their model?
- Google is changing and now influences your searches
- CDC has data on real occurances
- doctor / hospital visit where the flu is noted / tested for
- 10 regions - multiple locations per region - 300 clinical labs
- also people who died
- Google has information on how much people are wondering about the flu
- knowledge of vaccines and herd immunity (CDC has records)
How predictions are made from the data
- not sure about google (other than correlated search terms)
- CDC makes predictions based on actual medical data and trends
Search terms that correlate
- just because a term is correlated doesn’t mean they are related
- things influencing the flu are different than things influencing searches