Knowledge representation and acquisition approach based on decision tree

The knowledge representation and acquisition (KRA) is always a bottleneck problem of building artificial intelligence system, which is based on knowledge. We analyze the shortage of the KRA methods at present and proposes a new KRA method based on the decision tree (DT). A decision tree represents expert knowledge by its nodes, branches and leave, thus the knowledge acquisition problem can be converted into the learning problem of decision tree. In the process of building decision tree, we propose a new DT learning algorithm: rough-IDS algorithm, which is based on the rough sets theory and information entropy theory. With this algorithm, the decision tree can be simplified and its classified capability is improved. The instance analysis shows hat the proposed approach can represent and acquire the expert knowledge very well and provide a new approach for the expert knowledge representation and acquisition.