Business Rule Development

[MIDWESTJJS]

Looking for comments regarding structured business rule development and maintenance. Has anyone used commercial applications or used data mining techniques in extracting business rules from legacy systems.

[pbestgen]

I use the Data Mining "A Priori" algoritm to discover Business Rules on the fly. I work with WEKA (an open source software from Pentaho BI Suite). The Idea is the following. If you consider that your data are in most cases "good data", you could use a sampling of your records (e.g. 50.000 records) and let A Priori discovering hidden relationships between fields.
Some rules discovered have real business meaning but in most case they don't :

e.g. a rule like : if MARRIED_FLAG = Y then AGE > 18 (prob 87.5%), have a business meaning
but a rule like this : if MARRIED_FLAG = Y and GENDER = M then COUNTRY = FR (prob 100%) is not usefull if you all your clients are French for exemple.
So you have to filter the huge amount of rules discovered by A Priory.

In my case I keep every rule in a knowledge base and use them with a data quality tool in order to give a global data quality score (0% to 100%) to a data set. This could be usefull for a first global assesment of your data (does my data set looks good or bad ?).