带禁忌约束的频繁模式并行挖掘研究  

Research on Parallel Mining Frequent Patterns with Taboo Constraints

作　　者：薛胜军[1,2] 赵洪昌[1] 

机构地区：[1]南京信息工程大学计算机与软件学院,南京210044 [2]武汉理工大学计算机学院,武汉430063

出　　处：《武汉理工大学学报》2013年第3期125-131,共7页Journal of Wuhan University of Technology

基　　金：国家自然科学基金(41275116)

摘　　要：频繁模式挖掘技术在关联规则发现方面运用得十分普遍,已逐渐成为数据挖掘领域的研究热点之一。研究人员发现传统的频繁模式挖掘算法在挖掘过程中会产生大量的中间数据和用户不感兴趣的结果数据。这些数据在计算和存储方面的开销,对如今迅猛发展的海量数据挖掘来说无疑是一个挑战,严重影响了挖掘效率和准确性。针对这个问题,文章结合当下流行的Hadoop技术,对传统频繁模式挖掘算法进行分析和研究,提出一个带禁忌约束的频繁模式云挖掘算法模型。该算法模型利用Hadoop框架技术,对频繁模式挖掘过程中的模式长度和属性进行禁忌约束,分布并行地完成挖掘任务。实验结果显示,该算法模型在海量数据挖掘方面比传统算法更有优势。Data mining technology ba, sed on frequent pattern is widespread used on finding association rules and has gradually become one of the hot researc~ field of data mining. The researchers found that the traditional frequent pattern mining algorithm will produce a large number of intermediate data and the results which users are not interested in. These data in terms of computation and storage overhead is undoubtedly a challenge in rapid developing of massive data mining, and seriously affected the mining efficiency and accuracy. To address this problem, the paper combined with the current popular Hadoop technology has done some analysis and researches on traditional frequent pattern mining algorithms and has proposed a cloud data mining algorithm based on frequent pattern with the taboo constrained. The algorithm uses the Hadoop framework to restrain the length and attributes of the pattern in the frequent pattern data mining process and dis- tributed parallel complete the mining tasks. And the experimental results show that the algorithm has more advantages than the traditional algorithms in terms of mass data mining.

关 键 词：数据挖掘 频繁模式 HADOOP 禁忌约束 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]