贝叶斯网络预测平台的设计与开发

Design and development of Bayesian network forecasting platform

针对在实际应用中,需要根据不同的对象建立不同的贝叶斯网络来解决预测问题,设计并开发了贝叶斯网络预测平台,介绍了平台的结构和功能,重点介绍了平台实现时网络的数字化和网络拓扑结构的问题。利用数字形式描述网络的全部信息,用关系矩阵直观的描述节点间的依赖关系,并据此确定网络的拓扑结构,利用基于随机数的仿真算法对网络进行推理。该平台简单易用,为贝叶斯网络的建立和推理提供了一个通用的运行环境。

The fact that Bayesian network must be constructed for a specific object in the practical application motivated to design and develop a Bayesian network forecasting platform. The structure and function of the platform are discussed. The digitization and the topological order of network are importantly discussed. The number pattern is introduced to describe the topological model of network; relation matrix is adopted to represent those relationships among nodes, and the network topology is decided by this. Random based on sampling simulation arithmetic is used to reason the network. The platform provided an effective and flexible operation environment for construction and reasoning of Bayesian network.