一种面向流数据频繁项挖掘的降载策略  

Load-shedding strategy for data stream frequent item mining

作　　者：邹永贵 龚海平[2] 夏英 宋强[2] 

机构地区：[1]中韩合作重庆GIS研究所,重庆400065 [2]重庆邮电大学计算机科学与技术学院,重庆400065

出　　处：《计算机应用研究》2011年第4期1304-1307,共4页Application Research of Computers

基　　金：重庆市计算机网络与通信技术重点实验室开放基金资助项目(CY-CNCL-2009-01);重庆市科技计划资助项目(CSTC2009CB2015)

摘　　要：分析了目前降载技术的不足,提出了一种面向挖掘流数据频繁项集的降载策略。该策略采用了基于元组出现频率的语义删除策略,优先删除出现频率相对较低的元组,从而有效解决了在挖掘流数据中的频繁项所遇到系统超载时出现的问题,同时采用了根据流数据产生速率自动地控制是否启动降载策略,有效地解决了降载的适应性问题。通过实验和分析,证明了该策略在流数据频繁项挖掘中的有效性。This paper analyzed the deficiencies of current load shedding techniques and proposed a new load-shedding strategy for data stream frequent data item mining.This strategy adopted the semantics of tuple deletion based on data item frequency to delete tuples with relatively low frequency as possible,thus it could effectively solve the problems when mining the frequent data item while the system was overloaded.Moreover,starting and stopping load shedding strategy was controlled automatically based on the data stream rate,so it was effectively address the problem of load shedding adaptability.According to the experiments and analysis,the proposed strategy has the effectiveness of mining frequent items in data stream.

关 键 词：流数据 流数据管理系统 降载 频繁项 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]