Data Mining Technology Tracks Spread of H1N1 Flu
Google's NLP Algorithm Sifts User Searches to Find Flu Symptoms
Oct 24, 2009 Victoria Nicks
The Novel A strain of the H1N1 influenza virus, commonly referred to as the swine flu, has been spreading around the country and the world fast enough to be referred to as a pandemic. President Obama has declared that the swine flu is a national emergency. As the swine flu continues to spread, artificial intelligence tools, such as Google's Flu Trends map, are extremely useful.
Benefits of Google Flu Trends
During the spread of a pandemic disease, information on how and where the disease is spreading is valuable to health providers. Predicting where the disease is breaking out helps to ensure availability of treatment and medicine for patients. Google has developed a system for analyzing search queries for certain flu-related terms in order to determine whether or not the swine flu has spread to a certain area. This is accomplished through data mining, and natural language processing.
Natural Language Processing Application
Google analyzes search queries with natural language processing, and retains the search terms in association with IP addresses for 18 months. These search terms are typically used in behavioral advertising, in which advertisements are directed toward individuals based on their search history. The search information gathered for this purpose can also be used for other purposes, such as Google Flu Trends.
Analyzing the Spread of the Swine Flu
By scanning search queries for common flu symptoms with natural language processing, Google can determine whether there are increasing numbers of searches that may be due to actual flu symptoms. Associating those queries with the IP address of the source computer allows Google to associate the queries with a geographical location. The combination of these two types of information processing shows increasing or decreasing flu activity in different areas around the world.
Accuracy of Flu Spread Predictions
To verify the accuracy of Google Flu Trends, Google compares their projections against actual flu data provided by laboratories and hospitals around the world. Currently, Google Trends has only developed a system to identify the spread of influenza-type infections, but has been able to consistently predict the spread of the flu 1 to 2 weeks ahead of the CDC, which has to rely on doctor and hospital reports.
Saving Lives with AI Applications
While Google's information is incomplete and imperfect, this use of artificial intelligence technology may save lives as the swine flu pandemic spreads. Increasing supplies of antiviral medications in advance of an outbreak in an area allows health care providers to reduce complications in flu patients once the outbreak occurs. Preparation for swine flu outbreaks can only be accomplished with adequate information, which is a service provided by Google Flu Trends.
References:
Detecting influenza epidemics using search engine query data
Published February, 2009 by Jeremy Ginsberg, Matthew H. Mohebbi, Rajan S. Patel, Lynnette Brammer, Mark S. Smolinski & Larry Brilliant
