Products & Services: Model Builder

Model Builder is a development environment that accelerates the creation and deployment of sophisticated analytic models and is a pillar of Fair Isaac’s Enterprise Decision Management (EDM) technologies.  By enabling predictive modeling professionals to visualize and explore predictive patterns, build models, evaluate their quality, and swiftly deploy them, Model builder offers unprecedented power and control over the modeling process.  For the business, these models maximize the precision of every decision by informing them with insight into future outcomes and customer behaviors.

Designed from the modeler’s point of view, Model Builder covers the entire predictive modeling lifecycle, including development, deployment and continuous improvement.  Users work within a collaborative, object-oriented environment, facilitating reuse of models and modeling components across multiple projects.

The base Model Builder application provides several different modeling technologies including Linear Regression, Logistic Regression and Neural Networks.  Powerful Innovation Modules, leveraging advanced analytic capabilities developed by Fair Isaac can be added to the base Model Builder application.  Currently available offerings include the Scorecard, Segmentation ART™ and Data Spiders™ modules.

Model Builder Scorecard is a powerful and efficient predictive model development tool.  Model Builder Scorecard enables your modeling team to leverage the insight and innovation of Fair Isaac's industry-leading scorecard development methodologies.  The result is more powerful models that deliver greater "lift," which means greater profits for your business.

Model Builder Scorecard offers the same model development environment Fair Isaac analytics experts use every day to build powerful custom predictive models for our clients worldwide across industries, including financial services, insurance, telecommunications and healthcare.  Model Builder Scorecard provides a powerful set of advanced analytic techniques, such as Performance Inference, Multiple Fitting Objective Functions and Bootstrap Validation.

In addition, Model Builder Scorecard contains productivity enhancement features, such as Automated Variable Binning, that automate many of the routine, though important, tasks undertaken by your modeling staff.  Also within Model Builder Scorecard is the capability to share the work of your modeling team within a Model Repository. The net result is your team is able to bring your models into production more quickly and easily.

Model Builder Scorecard is a key component of our Enterprise Decision Management suite of software and services.  These help you to make smarter, more efficient and consistent customer decisions by leveraging Fair Isaac's industry-leading capabilities with our Fair Isaac Blaze Advisor rules management system, Model Builder Scorecard predictive modeling tool and Decision Optimizer strategy optimization software.

With Model Builder Scorecard, not only will your team be able to build predictive models-including characteristics, weights, reason codes and variable generation-but also deploy them for inline or batch execution in the Model Builder scoring server or directly within Blaze Advisor™ business rule management system flows.  The result is a more efficient deployment mechanism for your models that decreases your IT burden and gets your models into production more quickly.

Data Spiders module supports the creation of more powerful predictive models leveraging transaction-based variables.  By utilizing genetic algorithms, Data Spiders intelligently automates the search among thousands of potential transaction summaries to discover the most powerful collection of predictors, replacing the time consuming process of brainstorming, coding, testing and evaluating predictive summaries of transaction summaries.  Data Spiders improves the precision of your models while accelerating the data preparation process and allowing you to leverage valuable transaction data. 

Segmentation ART module provides Fair Isaac's proprietary Adaptive Random Tree (ART) algorithm for designing optimized segmentations of your portfolio.  Segmentation ART leverages genetic search techniques to automate the process of evaluating thousands of candidate segmentations and identifies the optimal segmentation for your portfolio.  Segmentation ART improves the precision of your models while significantly reducing the time required to develop effective segmented model systems. Learn more about Segmentation ART in the ViewPoints article "Leveraging Randomness: The Science of Adaptive Random Trees"

Like all of Model Builder innovation modules, Segmentation ART integrates seamlessly with the base Model Builder platform allowing you to leverage the capabilities of Model Builder for importing data, visualizing and exploring predictive patterns, defining predictive variables, and creating, evaluating, and deploying models

Model Builder for Decision Trees is a graphical development tool for rapidly creating decision strategies that combine data-driven analysis with human expertise.  Strategies are rendered as easy-to-view, editable decision trees.

• Explore extremely large datasets for insights into risk, profitability and other performance dimensions.
• Develop strategies that incorporate multiple objectives with alternating performance variables; also arbitrarily complex splits, data mapping and sample weighting.
• Use classification and regression trees to examine and split strategy segments to produce branches that isolate similar groups of data. 
• Automatically generate strategies-then explore and refine them interactively.
• Run sample data through trees to see the distributions.
• View profiles of strategy segments, run statistical analyses and explore the results of changing variables and values.
• Deploy strategy trees directly into Blaze Advisor, where they can be combined with predictive models and rules into decisioning applications and services. Blaze Advisor automatically generates the code required to execute model processes in almost any operational environment.  Alternatively, you can output your model as standard Java components from Model Builder Scorecard.