In a recent speech, Google CEO Larry Page claimed that big data analysis such as Google Flu Trends would be able to save up to 100,000 lives each year.
This promising outlook has now been put into question by a study conducted by the American Journal of Preventive Medicine. John Ayers, a researching professor at San Diego State University, even considers over-reliance on the results of such analysis to be potentially harmful to public health. Both the study of Ayers’ research team and other recently published papers on the topic prove Google Flu Trends’ failure to predict accurate numbers in most cases.
Using publicly available data provided by Google, manually giving differing weight to the monitored queries, as well as using artificial intelligence systems which automatically correct the bias of the queries, the research team was able to significantly improve the accuracy of such predictions. Google’s analysis overestimated the number of cases to a degree of about 73%, compared to about 26% from the upgraded system.
While Page suggested that an oversensitivity on privacy matters was a major obstacle for the effective analysis of big amounts of medical data, this study puts an emphasis on improving the way the collected data is analysed. Benjamin Althouse, another author of the study underlines the importance of transparency in such endeavours, stating that “Reproducibility and validation are keystones of the scientific method, and they should be at the centre of the big data revolution.”
While Google did not respond to the claims made by this specific study, they did respond to an earlier research, stating that they “welcome feedback on how we can refine Flu Trends” and that they “review the Flu Trends model each year“ in order to improve their accuracy. Ayers emphasised that it is not their goal to discredit the potential of big data in the field of public health, merely that it is necessary to create a platform for communal expertise to live up to it.
Read more here.
(image credit: Sari Dennise)
Interested in more content like this? Sign up to our newsletter, and you wont miss a thing!