CMNS 353 Lecture Notes - Lecture 10: Google Flu Trends, Data Collection, Star Wars Kid
16 Jul 2016
School
Department
Course
Professor
Document Summary
The parable of google flu: traps in big data analysis. By david lazer, ryan kennedy, gary king, alessandro vespignani. Feb 2013 gft made headlines for incorrectly predicting double the proportion of actual doctor visits for influenza like illnesses. Two issues that caused the gft mistake is big data hubris (over confidence) and algorithm dynamics. Big data hubris is the assumption that data is the substitute for traditional data collection and analysis. People suggested to throw out terms that seemed useless but this lead to gft completely missing 2009 flu epidemics. 2011-2012 gft missed a high of 100 out of 108 weeks starting in aug. 2011. Cdc data does a better job than the gft. The only way to heal gft & cdc errors is to combine both of their data. Gft was unstable due to google"s search algorithm. Algorithm dynamics are the changes made by engineers to improve commercial service and by consumers in using that service.
