At Statens pensjonskasse we use IBM/ILOG Jrules 6.7 for rather heavy pension calculations. Our largest application has over 1400 rules, a few decision tables and a about 100 ruleflows.
We have our own implementation of the IlrTool interface, monitoring rule execution during regression testing. We use this monitoring possibility, along with a few tricks to improve test coverage in our regression test suites. Hopefully I will cover this in a later post. This time we will jump straight to the eye candy.
I have hooked up the rule and package structure from our application for pension calculations with rule firing data from a couple of pension calculations. Using Gephi this gives some nice visuals in my opinion.
Each pink node is either a rule package(large) or single rule(small). The packages are hierarchical (connected with edges). A single pension calculation (blue node) triggers only a subset of the rules (yellow edges).
A more complex calculation...
These are hi-res images so feel free to click and open them in a separate window to get the details.
This is only two examples of rule firing patterns. For those of you that have worked with Norwegian pension law, you probably also got the feeling it is as many patterns as citizens in Norway…
