中文摘要：

贝叶斯网是不确定性知识表示和推理的有效框架,定性概率网是贝叶斯网的抽象表示,它简化了不确定性知识的表示,加速了不确定性知识的推理.近年来,定性概率网成为了不确定人工智能和知识发现领域的一个重要研究方向.分析总结了目前定性概率网的挑战和研究成果,主要包括定性概率网的知识表示、推理及应用3个方面.表示方面,概括了定性影响、定性加协作、定性乘协作和原因间影响,并探讨了定性概率网表示机制存在的问题及主要解决方法;推理方面,描述了推理算法,探讨了定性概率网推理冲突问题,分析比较了解决推理冲突的主要方法;应用方面,概括了基于定性概率网的知识发现和决策支持的主要方法.此外,基于对现有研究成果的分析总结,也指出了定性概率网相关方面进一步研究的问题及重点.

英文摘要：

Bayesian network（BN） is an effective framework for the representation and inference of uncertain knowledge.Qualitative probabilistic network（QPN） is an Abstraction of general BNs.By means of QPN,the representation of uncertainties is simplified and the efficiency of inference is improved.In recent years,QPN has become an important research issue in uncertain artificial intelligence and knowledge discovery.In this paper,we analyze and summarize the currently principal research findings of QPN,including its representation,inference and applications.As for the representation,we summarize qualitative influence,qualitative additive synergy,qualitative product synergy and intercausal influence.As well,the problems and representative solutions for QPN representation are discussed.As for the inference,we introduce the inference algorithm and discuss the problem of inference conflicts.We then mainly analyze and compare the representative methods for resolving conflicts.As for the application,we summarize the QPN-based ideas for knowledge discovery and decision support.In addition,based on the analysis and summary of existing research findings,we also point out some trends and emphasis of further study of QPN-related issues.