/fraud-editor

A scaffolded JSON editor for creating, testing and analyzing fraud rules

Primary LanguageJavaScriptMIT LicenseMIT

F.R.I.E.N.D

Fraud Rule Investigator, Engineer, Navigator and Deployer

GOAL

A scaffolded JSON editor for creating, testing and analyzing fraud rules.

Should be able to:

  1. Create rules by:
    • choosing from a bouquet of signals and operators
    • defining fail conditions
    • specifying if is a stoprule and whether all or any of its conditions must fail
    • nesting with additional rules as many levels as desired
    • schema should make it impossible to create illegal rules
  2. Test rules by:
    • providing a SQL query that produces an array of signal objects
  3. Analyze rules by replaying the rule historically:
    • specify date range
    • compute the number of false positive and false negatives
    • compute expected precision and marginal precision
    • compute expected revenue loss and loss reduction

NICE TO HAVES

  • Generate and/or use
  • Use for customer predictive analytics as well

Installation

Follow the npm download/install instructions as described in the JSONEDITOR project

Links