Rule graph: Incorporate expert and statistical knowledge for rule execution

We present an efficient graph based rule execution method which incorporate expert knowledge and statistical knowledge. How to use both the knowledge from expert experiences and the statistical information from business instances to improve the efficiency of rule execution is meaningful. We define a directed acyclic graph to control rule execution where each potential sequential rule execution corresponds one path in the graph. Expert knowledge is defined as the constraints on the executions of rules, and it can be used to prune the rule graph to reduce the potential paths. Statistical knowledge comes from the execution of a large number of business instances, and it can be used to assign different weights for paths and then adjust the structure of rule graph. Experiments indicate our approach can achieve a very efficient performance for rule executions which outperforms the existing approaches.