Learning the priority for rule execution

We present a novel method to learn the priorities of rules for sequential rule execution during the running of a rule engine system. The priority based ordering of rules influences the condition evaluation count of rule execution. User-assigned priorities can not guarantee optimal execution performance. We present and prove that the execution count and the dependency relationship are two factors influencing the optimal assignment of priorities. Furthermore, we propose an efficient priority learning algorithm where the priorities can be corrected during the running of the rule engine system to achieve a statistically minimal condition evaluation count. Experiments indicate that the execution performance of a rule engine system can be improved by using learned priorities.