摘要
分析了等价矩阵和联合决策矩阵规则提取算法对于大数据集低效性的根源.提出了基于任意分割的规则获取方法和相应的串行进位链计算流程.这种计算流程将大数据集上的规则获取,转化为通过分割后多个智能体(子系统)及其智能体间数据共享的"并行+串行"的规则提取计算过程,有效的解决了大数据集上规则获取问题.复杂度分析表明该算法在效率上较现有的算法有显著的提高;实例分析验证了该方法的可行性;相应的对比实验表明这种计算流程对大数据集上的规则获取的实用性和高效性.
Based on equivalence matrix and joint decision matrix,the reason of the existing algorithms inefficiency for rules extraction in massive data set is analyzed.The method of rules extraction and calculating process with serial carry chain based on the arbitrary division are presented.This process about the rules extraction will be changed into many agent(sub-systems) and inter-agent to share data by the "Parallel plus Serial" rule calculation,which can effectively improve the algorithm on the massive data set. Complexity analysis shows that the algorithm is more efficient than those existing algorithms. An example is used to illustrate the efficiency of the new algorithm. At last, experimental result shows that the calculating process with serial carry chain for rules extraction is not only efficient but also scalable.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2009年第12期2797-2802,共6页
Acta Electronica Sinica
基金
安徽省自然科学基金(No.070412061)