NGCI: Next Generation Communication Interoperability

Beyond Real Time - Prediction and Forecasting


In emergency management, it is important to not only know what is happening in the present, but also to forecast what might happen in the future. This can be done by using past data to model and extrapolate future trends and explore what-if scenarios. While the accuracy of these predictions is not guaranteed, it is useful to know where situations may be heading or where new situations may emerge. This allows the emergency management team to be more proactive instead of purely reactionary.

Micro-blogging sites such as Twitter have gained renown in their ability to gather voluntarily supplied updates on conditions and events during an emergency. The use of Twitter for such events has been studied, although much prior work has been using tweets in real-time or doing post-event analysis. However, real time analytics could easily be extended to forecast future topic trends (with varying degrees of certainty). ThemeRiver-type visualizations can be combined with topic extraction to visualize past and current tweet content, which could then be extrapolated to future points in time.

In addition to gaining situational awareness and predicting new trends, traditional forecasting domains should not be forgotten. Weather forecasts play an important role in emergency management and response and should be integrated as appropriate.

Although systems can be automated, care should be taken to ensure that manual controls can be applied in order to account for unforeseen circumstances.

The development of robust and reliable forecasting techniques can allow emergency management teams to plan ahead and proactively handle more of the emergency than would be permitted from just looking at real-time data.

Questions for the Community: 
What types of data do you currently forecast?
What types of data would you like to use in predictive models for emergency management purposes?


Subscribe to RSS - Beyond Real Time - Prediction and Forecasting