可变精度邻域区间值决策表的属性约简

Attribute reduction of interval-valued neighborhood decision table with variable precision

区间值决策表可以通过区间刻画对象相对于条件属性的取值,其在现实生活中运用非常广泛,该文在此基础上提出一种启发式约简算法。首先,为了提高信息系统的可辨识性,在区间值决策系统上引入可变精度阈值α与邻域阈值δ;随后,重新定义了决策表的上近似、下近似与正域;最后,定义属性质量度,以判定是否将条件纳入约简集合。为了更好地理解该算法的执行过程,该文进行了案例分析,并在4个数据集上完成了实验。实验结果表明,该算法具有良好的可行性,且其在准确率上优于另2种算法。

Interval valued decision tables are widely used for various fields in reality.This paper proposes a heuristic reduction algorithm for this kind of decision tables.Firstly,the variable accuracy thresholdαand the neighborhood thresholdδare defined on the interval-valued decision table.Then,this paper redefines the upper approximation,lower approximation and positive domain of decision table.Finally,attribute quality is defined to determine whether conditions should be included in the reduction set.In order to better understand the execution process of the algorithm,this paper conducts a case analysis and completes the experiment on four datasets.Experimental results show that the proposed algorithm has good feasibility,and its accuracy is better than the other two algorithms.