不完备系统中先验概率优势关系粗集模型及其数据挖掘方法研究

发布时间：2018-08-06 18:03

【摘要】：粗糙集理论的出发点是根据现有的知识对未知信息进行划分,然后确定每一个划分类对某一概念的支持程度,并用正域、负域和边界域这三个近似集合来表示,之后,通过属性约简和属性值约简算法获取决策规则。本文首先提出一种基于条件先验概率优势关系的粗糙集模型,此模型是建立在对不完备偏序关系决策系统属性值数据统计基础上的,既考虑到同一属性取值的不同情况又考虑到不同属性之间的关联性,使得各种先验信息能够充分利用,因此有效地提高了分类精度和分类质量。其次,由于基于知识粒度的不确定性度量方法不能精确、系统地反映系统的不确定性,为此本文提出一种新的基于边界域和知识粒度的改进粗糙熵。改进粗糙熵不仅考虑到因划分不精确所产生的不确定性,而且顾及到了由边界域的变化所带来的不确定性,从而使不确定性度量值的计算更加精确,为条件先验概率优势关系模型中不确定性度量问题的研究开拓了新的思路。最后,本文介绍了约简、分布约简和分配约简,并详细分析了三者之间的关系和它们的性质。同时,提出了基于改进粗糙熵的启发式约简算法和基于目标分配矩阵的分配约简算法。理论分析表明,后者因在求取约简过程过于繁琐而降低了搜索效率,而前者在约简过程中,直接删除系统中不必要的属性,因此节省了搜索时间,提高了搜索效率。
[Abstract]:The starting point of rough set theory is to divide the unknown information according to the existing knowledge, then determine the degree of support of each partition class to a certain concept, and express it with three approximate sets: positive domain, negative domain and boundary domain. Decision rules are obtained by attribute reduction and attribute value reduction. In this paper, a rough set model based on conditional priori probability dominance relation is proposed, which is based on the statistics of attribute values of incomplete partial order decision system. Considering not only the different values of the same attribute but also the correlation between different attributes, all kinds of prior information can be fully utilized, so that the classification accuracy and classification quality are improved effectively. Secondly, because the uncertainty measurement method based on knowledge granularity is not accurate, it systematically reflects the uncertainty of the system. Therefore, a new improved rough entropy based on boundary domain and knowledge granularity is proposed in this paper. The improved rough entropy takes into account not only the uncertainty caused by the inaccuracy of partition, but also the uncertainty caused by the change of boundary domain, which makes the calculation of uncertainty measure more accurate. It opens up a new idea for the study of uncertainty measurement in conditional priori probabilistic advantage relation model. Finally, this paper introduces the reduction, distribution reduction and distribution reduction, and analyzes the relations among them and their properties in detail. At the same time, a heuristic reduction algorithm based on improved rough entropy and an assignment reduction algorithm based on objective assignment matrix are proposed. The theoretical analysis shows that the latter reduces the search efficiency because the process of obtaining reduction is too cumbersome, while the former directly removes unnecessary attributes from the system in the process of reduction, so the search time is saved and the search efficiency is improved.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TP311.13

【参考文献】