eyeDES® Releases

eyeDES 1.0 - December 2012

First release of eyeDES decision engine containing all basic functionalities of the eyeDES server like configuration, shell, shell text client, shell terminal, process, scoring, reasoning, some input options, some output options, history persistence.

 

eyeDES 1.1 - June 2013

The main features that have been introduced in eyeDES 1.1 are:

  • Online model switch: allow the user to replace a detection model while scoring transactions.
  • Configuration snapshots: labelled commit and rollback capabilities have been added to the shell, allowing the user to save his setup before applying changes and allowing the user to rollback his changes to a former version of the configuration.
  • Scheduling and automation: allows the user to automate shell commands or shell command scripts, for instance to schedule a model replacement during lower activity periods.
  • Input & output configuration: several formatting options and protocol settings have been added to provide more flexibility in input and output. This includes time and number formatting, conditional output and error formatting.
  • Metrics: eyeDES collects data on a number of system operation and fraud detection metrics in real-time, and provides a query interface to retrieve this information from the system. These metrics can be used to create a dashboard visualizing system processing and fraud detection performance KPI’s.
  • Speed improvements: Record process speed has been improved.
  • PMML support: eyeDES now imports models specified in a PMML compliant file format.
  • Generic integration API: eyeDES is now usable through a generic Java API, allowing the user to integrate eyeDES in their own software.
  • Improved reasoning: next to reason codes and static textual reasons, eyeDES now provides dynamic reason texts that contain more context information about the suspicious transaction and the model.
  • Online documentation: most of the eyeDES configuration options have been documented in the software so that the end-user can see their usage from within the shell.

 

eyeDES 2.0 - February 2014

The main features that have been introduced in eyeDES 2.0 are:

  • Distributed deployment: allow the user to deploy eyeDES on several nodes to make process performances independent from each other while keeping a centralized control entry point.
  • Web application: provides a web interface to the user for monitoring purposes and model execution scheduling
  • Dashboard: displays metrics about the system in charts using the web interface
  • Simple model switch: makes the scheduling of model switch simpler by providing a web interface for this kind of operation
  • Web terminal: removes the need for specific client tool by providing a web interface for the terminal to users with sufficient privileges
  • Feature Analytics model file format: the import command now supports an eyeDES-specific file format that makes it possible to describe further specific details about a model to import

 

eyeDES 3.0 - September 2014

The main features that have been introduced in eyeDES 3.0 are:

  • Model import: variable definition catalog can now be defined in a JAR file that relies on library of functions.
  • Classification: new classification algorithms, new feature calculation functions, improved segment dispatching, extract classifier identity from classification result.
  • Metrics: feedback loop mechanism to compute cumulative metrics.

 

eyeDES 4.0 - December 2015

The main features that have been introduced in eyeDES 4.0 are:

  • Classification: new classification algorithms, new feature calculation functions.
  • Model retraining: having an already trained model, eyeDES can now build a new one using more recent labeled feedback data.