Four new analytic technologies recently patented by FICO are being integrated into solutions for cyber security, the Internet of Things (IoT), model governance, and optimization.
FICO’s patent for Collaborative Profiling provides an efficient way to recognize when a person’s behavior is changing, and to rank-order the significance of those behavioral changes across many individuals.
FICO’s method distills a history of events/transactions down to a limited number of behavioral “archetypes.” Individuals become described in continually adjusting archetype distributions that can distinguish normal anticipated behaviors from risky/abnormal changes in behavior. This technology can be used to monitor for cyber security threats and fraudulent payment activity. The deep insights of the archetypes also allow marketers to focus campaigns on the most relevant consumers.
The IoT is composed of billions of devices that emit vast amounts of data, often in free-form text messages. Organizations are trying to better leverage this massive, yet largely unstructured, set of data producers in more automated and effective ways. To do this, the log messages must be parsed efficiently so advanced analytics can be used to recognize anomalies and predict future events.
FICO earned a patent by inventing a way to statistically parse unstructured data and determine typical temporal sequences. The key to this method for IoT is fast-streaming self-learning analytics that run efficiently at a very large scale. This technology is intended to improve the management and security of IoT devices, which can be anything from cars to factory equipment to household appliances to cable TV boxes.
Model governance is a challenge for organizations that use predictive analytics for mission-critical tasks, such as making lending decisions. Models tend to lose their effectiveness over time. FICO’s patented Autoencoder Self-Diagnostic technology provides ongoing model validation to help ensure models in production environments are properly suited to future changes in data and behaviors.
Autoencoding is a class of neural network. Autoencoders learn to mimic input data through a representation of complex interactions between different latent factors. This can be used to understand data that moves away from what a model was originally trained on, or to identify segments of data that are underrepresented in a model.
As firms in every industry pursue decision optimization, FICO patented a new method of automating complex analytic tasks required to optimize marketing offers. FICO’s technology creates a dynamic system capable of self-learning the optimal strategy after a few iterations of a test-and-learn cycle. This will allow companies to respond more quickly to the challenge of changing market conditions and customer demands.
For a complete archive of more than 70,000 articles published since 1995 search the CardFlash.com library.