基于规则控制的区间参数优化方法及应用

发布时间：2018-11-02 09:25

【摘要】：区间概念格的参数优化问题不同于传统的参数优化问题。来源于区间概念格的区间参数,其取值直接影响着格结构的规模和稳定性,并对后续基于此挖掘出的关联规则,分类规则的数目和精度以及决策准则的效率产生着影响。目前人为主观地选取区间参数值有很大的不确定性和弊端,鉴于此设计区间参数优化模型。首先,数据量暴涨的现状会导致格结构中概念结点冗余,因此要进行概念格的压缩约简。结合形式背景下二元关系对与对象的近邻的定义提出了区间概念格的压缩算子,构建了基于压缩理论的区间参数优化模型,通过调节压缩度获取格结点冗余最少时的区间参数,并给出实例验证模型的有效性。其次,考虑到区间参数的改变只会改变部分概念结点和格结构,因此在之前的重建算法上提出了概念格的更新算法。结合基于区间概念格的关联规则提取算法,提出了基于关联规则的区间参数优化模型。实例分析表明了当区间参数取值接近[0.5,1]时,由此挖掘出的关联规则的数目适中且精度较高。再次,鉴于概念格本身对数据分类的特点,设计了基于区间概念格的分类规则提取算法。发现当区间参数改变时,概念的分类规则的数目和精度都会随之变化,由此提出基于分类规则的区间参数优化模型,通过控制规则的数目和精度,达到调节区间参数的目的。实例验证了模型的有效性。最后,站在三支决策空间的角度上,给出了基于三支决策空间的区间参数优化模型的应用,通过图书推荐案例讨论了区间参数的改变对决策准则的影响,进一步验证了区间参数的有效取值,并达成近似一致。
[Abstract]:The parameter optimization problem of interval concept lattice is different from the traditional parameter optimization problem. The values of interval parameters derived from interval concept lattices directly affect the scale and stability of lattice structures and have an impact on the number and accuracy of classification rules and the efficiency of decision criteria. At present, there is great uncertainty and disadvantage in choosing the interval parameter value subjectively, in view of this design interval parameter optimization model. Firstly, the situation of data explosion will lead to the redundancy of concept nodes in lattice structure, so the reduction of concept lattice should be carried out. Combined with the definition of the nearest neighbor of the object under the formal background, the contraction operator of the interval concept lattice is proposed, and the interval parameter optimization model based on the compression theory is constructed. The interval parameters when the lattice node is least redundant are obtained by adjusting the compression degree. An example is given to verify the validity of the model. Secondly, considering that the change of interval parameters will only change some concept nodes and lattice structures, an updating algorithm for concept lattices is proposed in the previous reconstruction algorithms. Combined with the algorithm of extracting association rules based on interval concept lattice, an interval parameter optimization model based on association rules is proposed. The analysis of an example shows that the number of association rules is moderate and the precision is high when the interval parameter is close to [0.5 ~ 1]. Thirdly, in view of the feature of concept lattice to data classification, a classification rule extraction algorithm based on interval concept lattice is designed. It is found that the number and precision of the classification rules of the concept will change when the interval parameters change. Therefore, an optimal model of interval parameters based on the classification rules is proposed, which can adjust the interval parameters by controlling the number and precision of the rules. An example is given to verify the validity of the model. Finally, from the angle of three decision spaces, the application of interval parameter optimization model based on three-branch decision space is given, and the influence of the change of interval parameters on the decision criteria is discussed through book recommendation cases. The effective values of the interval parameters are further verified, and the approximate agreement is reached.
【学位授予单位】：华北理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O153.1

【参考文献】