An advanced decisioning platform should be able to support several types of decisioning logic. For this reason, we created the ClaimFocus Solution Suite upon the ChoicePoint Claims Analytics Platform. Its decisioning development components include functionality for experts to create and to maintain models and definitions that are used in decisioning operation. The current components are designed to be used by statisticians, mathematicians, business users and system analysts. Our operational decisioning components include "engines" that incorporate decisioning models and definitions into day-to-day operations. Each engine has been optimized for performance so that the decisioning platform can take a request, calculate a decision and return the result quickly. The system does not require expensive hardware for many configurations and scales linearly when more processors are added to the configuration to handle the most demanding requirements.

Engine Overview

Predictive Scoring Engine
Our predictive scoring engine provides a robust environment for migrating predictive models into production quickly. The scoring engine is written in C++ to maximize scoring speed. The engine is capable of processing over a thousand transactions per second in batch and returning a response in milliseconds for real-time implementations on a low-end server (e.g., Wintel server with a single Pentium processor). The scoring engine operates on predictive models written in predictive modeling markup language (PMML), so a new model can be migrated into production simply by changing a configuration parameter. Based on the configuration file, a score can be based on an ensemble of models, including hybrid approaches (where different modeling techniques are combined) and dynamic model definitions (where different models are selected to determine the score). Most ChoicePoint solutions include several individual models to achieve greater accuracy.

Identity Matching Engine
The identity matching engine includes proprietary algorithms and processes for search strategies and computing match decisions. The engine includes callable routines with a well-documented API for match key building, searching and matching. The routines can be invoked locally or via a socket interface using TCP/IP. The routines include default definitions for more than 50 countries and can be customized for unusual matching requirements. Matching functions compare two records and provide a score from 1 to 100 and match decision (matched, not matched or undecided) based on default and/or user-defined matching criteria. The default criteria support matching for people, organizations, addresses and other identifiers.