摘要
给出了并行挖掘关联规则的形式化描述和并行挖掘的模型。在研究基于Aprior算法的各种并行实现如CD、DD、IDD和HD算法后,针对这些算法扩展性差以及负载不平衡的缺点,提出了在IDD和HD算法上使用Sidle调度策略,有效地解决了IDD和HD算法中非常重要的候选项目集在各个处理器节点之间的划分问题,尽可能使得各个节点负载平衡,从而提高算法的效率。
This paper details formal description and model for parallel data mining algorithms. Based on studying various Apriori algorithm based parallal realization such as CD,DD,IDD,HD et al, the paper puts forward on the Siddle scheduling policy in IDD and HD algorithm owing to poor scalability and unbalancing in these algorithm. Therefore, IDD and HD are improved by means of adopting the Sidle scheduling policy to solve the problem of load balance effectively.
出处
《现代计算机》
2006年第7期27-30,共4页
Modern Computer
基金
安徽省高校自然科学科研基金(编号:2005KJ051)
关键词
关联规则
并行算法
分布式
负载平衡
Association Rules
Parallel Algorithm
Distribute
Load Balancing
