Integrating multiple knowledge sources using decision tables

Most integration methods proposed in the past needed domain experts to intervene during integration to resolve conflicts and contradictions. The integration time is thus very long due to the dialogue of the experts. We propose a genetic knowledge-integration method that automatically combines multiple decision tables into one integrated decision table. Our knowledge integration consists of four processes: translation, encoding, integration and post-processing. The translation process transforms each knowledge source into a uniform syntactical decision table. The encoding process encodes each decision table into a bit-string structure. The integration process chooses bit-string decision tables for "mating", gradually creating good offspring decision tables. The post-processing process further simplifies the decision table integrated. Our knowledge integration approach can effectively integrate multiple knowledge sources in an environment with good communication facilities of networks. This saves much time since experts may be geographically dispersed, and their deliberations may be very time-consuming.