Decision is defined as selecting and collecting the best or better model from various plans or strategies in order to attain some special objects, nowadays, more and more people are attached to decision support system .which has become one of the hottest research problems, there are great amount of induction models and reference algorithm, among them bayesian network is an important indeterminable model, however , the computing overhead of bayesian network is inefficient in concrete problems can be solved more efficient by using other models.In order to put bayesian models into practice, this paper puts forth a new model based on bayesian network according to prediction problems, called as E-D(evidence and destination) model, which redefines bayesian network through classing it's node into 2 kind of types and denoting it's node with various states(in bayesian network, a node always has 2 states). In our E-D models, we can map the actual problems to evidence or destination nodes to establish the decision models, as will simplify the bayesian model to some extent, but can express actual problem more efficient and available, at the same time , the author develops the reference algorithm of E-D model using Java multi-thread technology, experiences show E-D models and it's reference algorithm run faster than otheralgorithms and are suitable to solve complex network.At the begin of my paper, I illustrate some rudiment conceptions and knowledge about DSS. then , we discuss the E-D models and it's multi-thread algorithm in detail and describe the framework of system design and the common DSS model's implement, at last, some actual applications or examples are given based on E-D models.
|