A probabilistic approach to streaming analytics

In a world of event filtering everything is typically a yes or a no, but we at Cogility decided to take a new approach by applying probabilities to our input events thus giving us outputs of yes, no, and everything in between. Incorporating probabilities into data introduces a whole realm of problems that need to be addressed if you want to produce meaningful results at a speed that is useful for real-time analytics. This talk will focus on how Cogynt, a zero-coding analytics platform, uses Flink to create user defined Bayesian networks that run in real-time and issues that arose during implementation.
You’ll learn:
- How to design probabilistic models in Cogynt
- How probabilistic models can be used to perform risk analysis
- How to maximize performance of your models in Flink.
- When you shouldn’t use probabilistic models
- Issues you can encounter when dealing with probabilistic data models and ways to address