基于规则提取的决策粗糙集阈值自适应算法

摘要

为了消除或者抑制由不合理先验知识带来的分类不精确问题,提出了一种基于规则提取的阈值自适应方法.以约简结果中属性的数量最小和相应决策规则的可信度最大为目标,给出一种自适应求阈值的算法.结合引力搜索算法,并利用决策粗糙集中的阈值为一特定离散取值时不会改变等价类的划分结果这一性质,对搜索空间离散化处理,然后给出基于智能算法的自适应求阈值算法,求得的阈值能够使得用户做出决策的属性约简结果最优.最后,通过UCI数据集验证方法的可行性.%The approximate classification of decision-theoretic rough set exits problems because of unreasonable priori knowledge.An adaptive learning parameters algorithm is proposed to solve the optimum problem.Firstly,under the decision rules which are based on the decision theoretic rough set and its attribute reduction,the number of essential attributes is minimal and the reliability of decision rules is maximum.Secondly,combining with gravity search algorithm,and discretizing the search space by using the property that the partition result of the equivalence class won't change when the threshold value of the rough set of decision is the specific discrete value,an adaptive learning parameters algorithm based on intelligence algorithm is proposed.A better reduction result can be gotten from that,which be able to satisfy decision maker's requirement.Finally,the feasibility of the method is verified by UCI dataset.