An algorithm for attribute extraction is proposed to meet the objective demand of produc-tion scheduling rule discovery for data set attribute reduction.Firstly,the characteristics of the pro-duction data are analyzed,and the attributes of production data are divided into several sets according to their importance and correlation.Then,the importance objective function is established to find the important attributes by using the fuzzy entropy and the clustering accuracy.Finally,the correlation analysis is used to find the related attributes of the important attribute,which are then merged to form the important composite attribute to enhance the effect of attribute extraction.In order to verify the validity of the technology,a subset obtained by the technique is compared with another subset ob-tained by the stochastic method,and the compatibility and the accuracy of rule extraction between them are analyzed.The experimental results show that the data subset formed by attribute extraction has lower incompatibility and can concentrate the scheduling rule knowledge of the original data sets, which mean that the accuracy and efficiency of a variety of scheduling rule discovery algorithms can be improved significantly.Thus,the technology developed is suitable for the attribute extraction in the preprocessing stage of the production scheduling rule discovery.%针对生产调度规则提取工作对数据集属性约简的客观需求，提出了一种关键属性提取技术。首先，分析了生产数据的特点，并依据重要性和关联性，将生产数据的属性划分为多个集合；然后，在此基础上利用模糊熵与聚类准确度建立重要性目标函数，用于发现重要属性。最后，利用关联性分析查找重要属性的关联属性，将相关属性进行合并，形成重要复合属性，以进一步增强属性提取效果。为了验证该技术的有效性，将利用该技术所获取的数据子集与通过随机法所得到的数据子集进行了对比，分析比较了各数据子集的相容性和规则提取准确性。结果表明，提取属性后所形成的数据子集具有较低不相容度，浓缩了原始数据集的调度规则知识，可显著提升多种生产调度规则挖掘算法的准确度与效率。该技术非常适用于生产调度规则挖掘数据预处理阶段的关键属性提取工作。
展开▼
