The Power of 10: FICO’s Patent Powerhouse Unveils 10 New Patents

I am proud to announce the latest group of ten data science and software patents awarded to FICO by the U.S. Patent and Trademark Office. Our global data science and software teams are always working hard to be first to market with industry-focused, innovative technology that makes a difference, improving our customers’ business outcomes. Patents awarded based on the strength of our applications demonstrate the novelty of these innovations.

Applying Innovation to Meet Real-world Process and Technology Challenges

FICO’s latest patent wins illustrate the difference that a focused research process can deliver, applying FICO’s data science brawn to solve customers’ most pressing business and technology challenges. FICO’s new data science and software patents, which I collectively characterize as “the power of 10,” amplify the benefits that customers can achieve when applying FICO® Platform and our entire family of software solutions to run their business. I’m proud to be inventor or co-inventor on many of these data science patents, which bring my personal patent count to more than 130 patents authored, with 92 granted and 42 pending. Congratulations to all FICO inventors, the data scientists and software engineers who worked together to create these first-of-kind technology innovations.

FICO’s new data science patents offer a diverse range of business applications, empowering customers to drive innovation across machine learning, technology and process. The ten data science and software patents are:

• “Auto-Encoder Enhanced Self-Diagnostic Components for Model Monitoring” — which is a diagnostic technology system for model governance that provides a reliable indication on model degradation and recommendation for model rebuild. When applied to a class of supervised models, the diagnostic system can determine the most appropriate model for the client based on the reconstruction error of a trained auto-encoder for each associated model.
• “Overly Optimistic Data Patterns And Learned Adversarial Latent Features” — which enhances AI machine learning models to anticipate, detect, and mitigate adversarial AI attacks on related computing systems, models, and technologies.
• “Method and System for Predicting Adherence to a Treatment” — which characterizes an individual using predictive modeling techniques driven by patient treatment data to determine whether a patient is likely to comply with the treatment specified by a physician.
• “Supervised Machine Learning-Based Modeling of Sensitivities to Potential Disruptions” — which relates to developing and using machine learning-based models for quantifying sensitivity of an entity's expected performance to some future potential disruptions in ecosystem.
• “Computer-Implemented Decision Management Systems and Methods” — which relates to systems and methods that identify inter-relationships between various decision assets and provide relevant alerts and notifications where a local decision asset may globally affect other decision assets in different contexts.
• “Fast Automatic Explanation of Scored Observations” — which generates concise explanations of scored observations that compute efficient trade-offs between rank-ordering performance and explainability using the framework of partial dependence functions, multi-layered neural networks, and latent explanation neural network scoring.
• “User Interface to Analyze and Navigate Through Decision Logic” — which covers computer-implemented methods to efficiently analyze and navigate through decision logic using an execution graph.
• “Building Resilient Models to Address Dynamic Customer Data Use Rights” — which describes how to build resilient models that can account for events that can affect the composition or availability of customer data used in model development (e.g., loss of data rights, withdrawal of customer consent) without having to resort to model retraining. 
• “Similarity Sharding” — which involves a system that organizes data into shards tailored to similar relationships among data items, promoting more efficient management and preservation of related information.
• “Multi-Layered Self-Calibrating Analytics” — which improves fraud detection leveraging multi-layered, self-calibrating analytics for detecting self-learned latent features for detection of fraud in the absence of quality historical data.

Keep up with FICO’s fast-paced innovation machine by following me on the FICO Blog, LinkedIn and X, and I’ll see you in the New Year. Happy Holidays!

Explore FICO Patents in the U.S. and Other Countries

• Explore Previous Awarded Patents
• Learn more about our patents
• Discover the power of FICO Platform