The Research Of Bayesion Networks And Its Application In Case-Based Reasoning

The Bayesian Network(BN) proposed by Pearl is a new mechanism for uncertain knowledge representation based on probability theory and graph theory. BN is network structure with clarity semantics.lt exploits the structure of the domain to allow a compact representation of complex joint probability distribution.Its sound probabilistic semantics, explicit encoding of relevance relationships, inference algorithms and learning algorithms that are fairly efficient and effective in pratice, and decision-making mechanism of facility, have led BN to enter the Artificial Intelligence(AI) mainstream.For the reasons that they have produced more and more practical values and economic profits in many important application fields, such as modern expert systems, diagnosis engines, decision support systems, and data mining systems, researchers from both industry and academia are thus taking them much seriously.The Case-Based Reasoning(CBR) that has developed from the AI in recent about ten years is different from Rule-Based Reasoning(RBR). CBR systems are designed to match cases stored in a database with new cases. In other words, CBR systems use past cases and experiences as a basis for dealing with novel problem, evaluating the proposed solution, explaining abnormal thing or understanding new thing. The chief reason of the development of the Case-Based Reasoning is that there are lots of difficulties to obtain knowledge in the traditional Rule-Based Reasoning and that CBR systems don' t remember the solved problems, which result in low efficiency of reasoning, disability to deal with abnormal things and the rather friability of the system ability. While CBR systems can solve the above problems. CBR systems have lots of strongpoints, such as the completely expressing of the information, the incrementally learning, the precisely simulating of the visualized thought, the conveniency of the obtaining knowledge,the high efficiency of solving new things and so on.The thesis which chiefly researches BN,CBR and the applications of BN in CBR consists of six chapters. In the first chapter, the thesis illustrates the foundation and significance of this thesis and simplysummarizes their researchful history and actualities of BN and CBR.In the second chapter, the thesis firstly explains the notion of BN, afterwards studies the application of BN in Data-Mining (DM) in detail and also studies the learning of the probability parameter and the structuring framwork of BN in the condition of the full data and the lacked data. In the third chapter,the thesis discusses the basic principle and the constructing measure of the Na'ive Bayes, the enganced Naive Bayes and the current Naive Bayes and compares their advantages, disadvantages and applicable scopes. In the fourth chapter, the thesis introduces some primary concepts and principles of CBR, investigates and discusses the key technology of CBR. In the fifth chapter, We studies the application of the method of BN in the Retrieval and Maintenance of CBR. The last chapter is the summary of the whole thesis and the prospect of my research.