How Fuzzy Association Rule Mining and Twitter Can Save Lives
As the U.S. recovers from one of the worst cold and flu seasons in recent years, incidents of the avian flu are appearing across China, and other infectious diseases continue to spread across most regions of the world. Outbreaks like these can have immediate and negative impacts on our communities and put a strain on our health and emergency response resources.
Two recent innovations by the Johns Hopkins University (JHU) are using more sophisticated data modeling and social media technologies to improve the tracking and predicting of infectious diseases.
- A team of scientists from the JHU Applied Physics Laboratory has developed a way of accurately predicting outbreaks several weeks before they occur using a new data modeling tool (PRISM—Predicting Infectious Disease Scalable Model). PRISM uses Fuzzy Association Rule Mining (FARM) to extract relationships between multiple variables in a data set, such as clinical, meteorological, climatic, and socio-political data. The model can help decision-makers assess the future risk of a disease outbreak occurring in a specific area at a specific time.
- Social media has been used to track information about the spread of certain diseases for years. New tweet screening methods using sophisticated statistical modeling based on human language processing technologies developed by researchers at JHU are improving on this capability. They can filter out online chatter not linked to the actual disease to achieve more accurate results that more closely match government disease data and in much less time (CDC data can take up to two weeks to publish while social media is near real time).
These and other innovations enable communities to make informed decisions about limited resources and quickly anticipate the needs for outreach and health services to decrease the impact of future outbreaks.