A Compact Evidence Model for Knowledge Aggregation of Large-Scale Distributed Databases

When evidence theory is used to aggregate large-scale distributed databases, it is often the case that the evidence model has complicated BPA functions. Such an intractable model severely limits the application prospects of evidence theory in this area. In this paper, a compact evidence model is developed to represent uncertain knowledge discovered from large-scale databases. This model categorizes the uncertainty into two different levels and model them with a simple evidence and a reliability measure respectively, and then combined these two kinds of uncertainty together by a reliability discounting to obtain a united evidence model. The final united model may be intricate in form but it can be compactly represented by the simple evidence and reliability measure. This compact model is relatively easy to be handled by computer or be conveyed by network since it has only a few model parameters. Sometimes it can also reduce the computation loads needed for evidence combination greatly.