Record-Couple Based Production Rule Mining Algorithm: Tested in Medical Billing Domain
Umair Abdullah, Ahmed Aftab, Sohail Asghar, Kashif Zafar · 2014
Most research in the area of rule mining tends to focus mining of either ‘classification rules’ or ‘association rules’, while ignoring ‘production rules’. Production rules are building blocks of rule based expert systems. This paper presents an algorithm for production rule mining. Algorithm matches two states of an object; current state and previous state collectively termed as ‘record-couple’. The algorithm tries to find differences within a record-couple, and generates production rules from these differences. The proposed algorithm has been tested on synthetic data and then applied on actual data. The Algorithm is being used for supplying production rules to a rule based expert system which is applied in medical billing domain. The system is being used for finding knowledge oriented data inconsistencies in medical claims. Implementation of the algorithm is a demonstration of how data mining technology can be used for providing continuous supply of knowledge to rule-based expert systems.