The technology under the hood.
At Cogynt’s core is a specialized processing technology called HCEP (Hierarchical Complex Event Processing) — a technique that combines data models for multi-dimensional analysis at different complexity levels.
Cogynt is highly scalable. It leverages Apache Kafka as the messaging and persistence backbone, and Apache Flink as its stateful compute engine. Cogynt effectively integrates these open source technologies, making them transparent to the end user.
What is HCEP?
HCEP is both a top-down hypothesis-driven pattern definition process and a bottoms-up event matching and mapping process.
The principle benefit of HCEP over the simple version of CEP is that it allows the analyst to address more difficult problems in an integrated context. For example, if we are looking at human behavior, there can be many factors that go into a profile, such as spending patterns, criminal violations, social media usage, personal associations and travel. Any one of these factors alone might have limited meaning, but, in combination, they could represent a potential risk.
HCEP looks at non-events as well, meaning that a pattern can “look” for events that have not yet happened, and, in some instances, that non-occurrence can be just as important as an occurring event itself.
HCEP also continuously maintains the state of the hierarchy. If for some reason the system labeled a person’s profile as being high risk, and new data came in contravening that assessment, that person’s profile would be automatically adjusted.
Elements of Cogynt
Computational Hierarchy
A computational hierarchy is the arrangement of patterns that form a hierarchy, where patterns at lower levels feed higher level patterns, allowing an increase in abstraction and therefore understanding as patterns are populated up the hierarchy.Stateful CEP
Stateful CEP persists information about the state of ALL pending patterns allowing for incremental evaluation of CEP patterns and to respond to new information (events) in the shortest time, while continuing to track potential patterns over a very long period of time. The persistence of this provides the means to “understand” the entire field of endeavor continuously.Bayesian Belief Network (BBN)
BBN network determines the statistical likelihood of a future event. This is accomplished by applying Bayesian computations within the hierarchy and propagating those computations up the hierarchical chain determining the likelihood of future events at each level within the hierarchy.Domain Specific Language
A DSL that defines the logic of hierarchical patterns and the complex event processing. Our DSL is called Event Pattern Constraint Language (EPCL) and it is declarative and zero coding.
Zero Code. Low Touch.
The Cogynt Authoring Tool is completely graphical, requires zero coding, provides the ability to define hierarchical patterns using the EPCL, and maps the source data (Kafka topic schemas) to the patterns.
The Cogynt Authoring Tool performs basic consistency checks prior to deployment, ensuring the HCEP model is logically consistent and will execute once deployed.
The Cogynt Runtime converts the deployed HCEP model into the Flink model representation, a Directed Acyclic Graph (DAG), from which Flink processes all events according to the HCEP pattern design. All HCEP generated events are published back into Kafka for reingestion into higher-level patterns or can be published for visualization and further analysis.
Resources
- Video
- White Paper
- Blog
Get Started
Ready to learn more? Contact us today!
