A scaffolded JSON editor for creating, testing and analyzing fraud rules.
- 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
- Test rules by:
- providing a SQL query that produces an array of signal objects
- 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
Follow the npm
download/install instructions as described in the JSONEDITOR project
- Version based on josdejong's JSONEDITOR
- Preferred version based on Jeremy Dorn's json-editor
- Can specify a schema interactively that generates a form
- github
- [demo]