【作者】 林威；

【导师】 郭朝珍；

【作者基本信息】 福州大学 ， 计算机应用技术， 2006， 硕士

【摘要】 地震预报的实践与理论表明,地震孕育和发生是一个极其复杂的过程,存在很多错综复杂、关联耦合的相互关系,并存在大量的不确定因素以及不确定信息,有效地进行地震预报面临着诸多困难。因此,寻求能对复杂地震前兆异常相关的各种信息进行快速融合,并有效地处理不确定性知识的决策模型及方法,一直是研究者们不懈努力的方向。本文针对地震综合预报,特别是网络会商式地震综合预报群决策支持系统中广泛存在的不确定性问题和多源信息表示与融合问题,深入研究地震综合预报过程中贝叶斯网络模型的构造方法,并将其应用于网络会商式地震综合预报群决策支持系统。主要的研究如下: 1.在分析地震综合预报所面临的主要问题,总结现有地震综合预报模型与方法存在的局限的基础上,论述了应用贝叶斯网络方法进行地震综合预报决策方法的优势,并明确了地震综合预报贝叶斯网络需要研究与解决的主要问题。2.针对基于贝叶斯网络决策方法的“瓶颈”问题—贝叶斯网络模型的构造,本文进行了深入研究。引入面向对象的知识表示方法,建立了面向地震综合预报贝叶斯网络的知识表示体系,提出了基于地震综合预报学科专家知识的贝叶斯网络分级构造方法,为复杂的地震综合预报贝叶斯网络模型构造提供了一种有效的知识表示策略。给出了一种分级层次的地震综合预报贝叶斯网络模型,为复杂地震综合预报贝叶斯网络模型的构造提供了系统的指导原则。3.本文深入研究了贝叶斯网络理论在地震综合预报中的应用,针对贝叶斯网络理论与地震综合预报自身的特点,给出了地震综合预报贝叶斯网络模型在地震综合预报,特别是网络会商式地震综合预报群决策支持系统中的应用策略。建立地震综合预报贝叶斯网络分析器,以可视化方式构造与修改贝叶斯网络。在深入研究标准桶消元算法的基础上,采用引进消息传递机制的桶树消元算法作为网络推理模块,并对其进行了改进。总之,本文深入研究了贝叶斯网络理论在地震综合预报中的应用,形成了较为完备的地震综合预报贝叶斯网络模型,并在网络会商式地震综合预报群决策支持系统中获得成功应用,为解决地震综合预报中的不确定性问题和多源信息表示与融合问题提供了一条切实可行和卓有成效的途径。更多还原

【Abstract】 Both of seismological practice and theory indicate that the generation of earthquake is an extremely complicated process. With so much anfractuous and correlation, many uncertainty factors and much uncertainty information in complex coupling precursory anomalies, the earthquake prediction is very difficult. The researchers in this field always try their best in investigating decision models, which can quickly fuse precursory anomalies related to comprehensive earthquake prediction, and effectively handle the uncertainty information. This dissertation deeply investigates several key techniques including knowledge expression of Bayesian Networks, model construction of Bayesian Networks, and Bayesian Networks update algorithms. Also it presents the designs and implemental methods of Bayesian networks application for comprehensive earthquake prediction, especially Network Negotiatory Comprehensive Earthquake Prediction GDSS. The main points of research can be summarized as follows. 1. The main difficulties faced by comprehensive earthquake prediction are analyzed, and the major limits of existed comprehensive earthquake prediction decision models and methods are summarized. Then the Bayesian network application method for comprehensive earthquake prediction is put forward, and several advantages and the principal goals of this method are pointed out. 2. In order to handle the construction and reasoning problems, a hierarchy construction method for Comprehensive Earthquake Prediction Bayesian Networks Model is established, which is based on seismological experts’ experiences and knowledge, and it also provides a systemic principle for Comprehensive Earthquake Prediction Bayesian Networks Model construction, which deals with complex precursory anomalies. 3. This dissertation deeply dose researches about implement methods of Bayesian networks application for comprehensive earthquake prediction, especially Network Negotiatory Comprehensive Earthquake Prediction GDSS. An analyzer for Comprehensive Earthquake Prediction Bayesian Networks Model is developed, which can construct and modify Bayesian networks through visual interfaces, and its inference module is based on improved bucket tree eliminating algorithm. In conclusion, this paper deeply researches the applications of Bayesian networks for comprehensive earthquake prediction, proposes a Comprehensive Earthquake Prediction Bayesian Networks Model, which was successfully applied in the Network Negotiatory Comprehensive Earthquake Prediction GDSS.更多还原