复杂系统预测中知识约简算法及其表示的研究

【作者】 梁泉；

【导师】 陈宇拓；

【作者基本信息】 中南林学院 ， 计算机应用， 2004， 硕士

【摘要】 本论文针对复杂不确定性系统的特性，介绍了将知识发现理论方法与传统的预测理论方法有机地结合起来进行预测研究的新方法。针对复杂不确定性系统预测中的重要问题：不完整、不精确知识的发现利挖掘，利用粗糙集理论，提出了知识属性约简和值约简的简化算法，降低了属性约简和值约简算法的运算量和难度，为同类算法研究提供了一个很好的研究方向。对于分类知识的表示，提出了一个能降低概念格构造复杂度的简化算法，具有一定的理论研究价值。主要研究工作包括： (1) 对复杂不确定性系统的特性及其传统预测理论方法进行了研究，介绍了将知识发现理论方法与传统的预测理论方法结合起来进行预测研究的一些理论，并介绍了目前的一些研究情况。 (2) 针对传统统计学、概率论等预测方法不适用于海量数据预测的特点，研究了基于粗糙集理论的分类知识的挖掘算法。本论文利用粗糙集理论主要研究不完整、不精确知识且不需要先验知识等附加信息的特性，研究了粗糙集的基础知识、分类知识的挖掘算法以及有关的属性约简、值约简算法，并提出了简化算法，降低了算法的复杂度，消除了挖掘知识库中知识冗余。 (3) 在粗糙集理论与挖掘算法的基础之上，研究了分类知识的概念格表示方法，介绍了概念格的理论与知识表达方法、具有完备性的分类知识的挖掘算法，重点研究了基于概念格的分类知识表达结构及其构建算法，并在认知心理学与信息熵的理论基础上，提出了构建概念格的简化算法，降低了概念格构造复杂度。更多还原

【Abstract】 Aiming at the properties of complex uncertainty system, the paper has introduced the new methods which organically combined the theory and method of knowledge discovery with those of the traditional forecast. Aiming a important aspect of forecast of complex uncertainty system-to disvover and mine incomplete&inaccurate knowledge,the author has proposed thesimplifying algorithms of attribute and value reduction based on rough sets. These two kinds ofalgorithms reduced the complexity and operation quantity,and provided a good direction ofstudy for relative research. For representation of classification knowledge,the author hasproposed a simplifying algorithm which can ruduce the complexity degree of constructingconcept lattice,and this has some theory value.Major research work in this paper includes:( 1) The author has studied the properties of complex uncertainty system and the theory and method of traditional forecast.The author has also studied the theories which organically combined the theory and method of knowledge discovery with those of the traditional forecast,and introduced some circumstances of studies at present.(2) Aiming at the characteristics of traditional statistics and the theory of probability inapplicable to forecast the large amount of data the author have studied the data mining algorithm of classification knowledge based on rough sets. According to the features of rough sets that mainly studies incomplete and inaccurate knowledge and need not added information such as prior knowledge, this author have studied basic knowledge based on rough sets,mining algorithm of classification knowledge relating with attribute and value reduction algorithms.After those studies,the author has proposed the simplifying algorithms of attribute and value reduction .These two kinds of algorithms reduced the complexity in conventional algorithm and eliminated the knowledge redundancy in the mined knowledge base.(3) On the basis of rough sets and data mining algorithm the author have studied the representation method of concept lattice for classification knowledge,and introduced the theory of concept lattice and method of knowledge representation, presented mining algorithm of complete classification knowledge.The authour’s studies focus on the structure of classification knowledge representation with its constructing algorithm. Based on cognition psychology and information entropy the author has proposed a simplifying algorithm that reduced theconstruction complexity degree.更多还原