IBM identified three big trends in predictive analytics that will drive some of its development efforts around predictive analytics:
- New Data
New data types such as images and video as well as new sources like RFID and mobile devices must be accounted for and become part of how predictive analytics are built. Access to this data is going to be important but so will techniques to rationalize this data as the nature of the data has changed. I blogged about IBM’s Big Data strategy which is clearly relevant to this. SPSS also has a history of breaking modeling with large datasets down into multiple smaller modeling tasks that can be distributed and combined for a final model that will also play into handling these new large, rapidly changing data sets.
- New Techniques
There is continual research into new predictive analytics techniques. In particular integration with business rules and optimization, new visualization techniques, the possible use of game theory, network analysis and evaluation approaches for outcomes must all be part of the solution. SPSS has already been doing some integration of these new techniques, as discussed in the social media session for instance.
Far too many models are still not deployed and used. Embedding predictive analytics into decision-making even as that becomes real-time and based on streaming data must be easier and more continuous. SPSS Decision Management shows clearly that this will remain a focus for SPSS.
Most of the rest of this session was under NDA but you can check out what I wrote about lessons from IBM’s Watson too as that seems relevant.
That’s a wrap for the IBM Business Analytics Summit 2011.