The capability to look around the corner and predict what might happen next based on what is happening now is an extension to the event pattern matching over time or event windows used in ED-SOA solutions.
An example is when Complex Event Processing is used to monitor real-time CDRs (Call Detail Records) streaming in a telecommunications network and identifies subscriber calls that are dropping in a short period of time. Once the event-driven system has determined that a problem is happening now for a specific subscriber number, the next step is to "look back" over past persisted data, which could be large, and evaluate whether this person is more or less likely to change his or her telephone company.
The analysis of the potentially vast amount of historical data creating models that use a collection of algorithms to predict behavior is generally handled by other technologies such as data mining or real-time decisions. However, Complex Event Processing platforms can leverage these models and then use them in real time.
The value in this use case would be that the person experiencing the dropped calls could immediately get a call from customer support or a SMS text message that could automatically offer an incentive to stay with his current supplier.