论文标题:非频繁关联规则挖掘的研究
Study of the Infrequent Association Rule Mining
论文作者 高法钦
论文导师 贾作皆,论文学位 硕士,论文专业 计算机软件与理论
论文单位 山东科技大学,点击次数 77,论文页数 85页File Size4229k
2003-05-01论文网 http://www.lw23.com/lunwen_278538537/ 数据挖掘;非频繁关联规则;牛顿插值;支持度;相似度;置信度
Data mining, infrequent association rule, Newton interpolation, support, similarity,confidence.
本论文研究数据挖掘的典型理论和一些急待研究、发展的课题，并将研究重点放在非频繁关联规则挖掘上。 简单的说，数据挖掘是从大量数据中提取或挖掘知识。 数据挖掘是计算机科学一个活跃的研究领域，出现于80年代后期，从多门学科发展起来，这些学科包括数据库技术、人工智能、统计学等。 关联规则挖掘是数据挖掘的一项重要功能，首先由Agrawal.等在1993年提出，用于发现大规模数据集中项集之间有趣的联系或规则。关联规则挖掘在分析零售店中消费者的购买模式、决策分析和商务管理方面是有用的。现在，关联规则挖掘已经成为理解数据的流行工具。 论文中介绍了关联规则挖掘的典型算法：Apriori算法、FP-G算法和多层关联规则挖掘算法。论文中还总结和介绍了关联规则挖掘近几年的一些重要的研究成果如运用多重相似查询技术提高数据挖掘效率、运用Apriori算法挖掘量化关联规则、基于集体度——置信度框架的关联规则挖掘、挖掘互依赖模式等。 关联规则挖掘面临着一些问题需要解决，本文研究了如下几个问题：非频繁关联规则挖掘、提高挖掘过程的交互性和自动化水平，每一个问题都给出了相应的算法模式或方法。在非频繁关联规则挖掘方面，本文重点研究了基于相似度——置信度的关联规则挖掘，并成功提出了一个算法模式(扩展算法)和一些相关理论。在提高挖掘过程的交互性和自动化水平方面，重点研究了如何用插值与逼近的方法自动确定支持度／置信度阈值，研究并阐述了将牛顿插值法用在关联规则挖掘中自动确定支持度／置信度阈值的方法和一些注意事项，并给出了详细的实验例子。
In this paper, I study the typical theory and some other problem that are waited to be solved about data mining and association rule mining, and put the important on the study of infrequent association rule mining.For simply,data mining is the nontrivial extraction or mining knowledge from large amount of data.Data mining is a major research domain of computer science since the latter 80s".It develops from many disciplines such as database technology, artificial intelligence and statistics,etc..Association rule mining is a major function of data mining.It is first introduced by Agrawal et.al in the year 1993 and used to discover the interest relation or rules between items of large data set. Association rule is used to analyze the consumer purchasing pattern in retail stores, tactics analyzing and business management and now, association rule mining have become a prevail tools of understanding data.We discuss some traditional algorithm of association rule mining: Apriori algorithm, FP-G algorithm and multi-layer association rule mining algorithm. We also summarize and introduce some achievement in the study of mining association rule in recent years, such as Using the technology of multiple similarity queries to enhance the efficiency of data mining. Mining the quantitative association rules with the improved Apriori algorithm, Mining association rules based on the framework of collective and confidence. Mining mutually dependent patterns, etc..Some problem that association rule mining confront are required to be solved. In this paper,we mainly study the problem: Infrequent association rule mining, Raising the capability of algorithm-user exchange and automatic level of mining progress, and give the corresponding algorithms or methods. In this paper,we put premium on the study of mining association rule based on the framework of similarity-confidence in the study of infrequent association rule mining, we give out the extension algorithm of mining similarity association rule and some relevant theory. In the subject of raising the automation level of mining progress, we mainly study how to use the technology of interpolation to determine the threshold of support/confidence automaticly, And give some detailed experimental examples.
|
