基于PrefixSpan的序列模式挖掘改进算法研究

发布时间：2018-06-07 10:36

本文选题：数据挖掘 + 序列模式　；参考：《四川师范大学》2017年硕士论文

【摘要】：序列模式挖掘是数据挖掘中一个重要研究领域,其序列模式挖掘算法研究具有重要研究意义。通过对经典序列模式挖掘算法的研究,本文取得如下研究成果:(1)对经典序列模式挖掘算法AprioriAll、GSP、SPADE和SPAM进行比较与分析。(2)对基于模式增长的序列模式挖掘算法Prefix Span进行研究,给出了该算法在序列模式挖掘过程中存在重复投影数据库的理论分析、算法在执行过程中的投影划分次数的上确界和下确界、最坏情形下产生重复投影数据库数的一般式。(3)在基于前缀投影思想并参考SPAM算法的数据结构基础上,采用二维表来存储频繁项在序列数据库中的位置信息;统计二维表中每列非空位置信息集的个数快速计算序列支持度,并确定其频繁性;通过对序列前缀位置的进行检查避免序列模式挖掘过程中对相同投影子表的重复挖掘。
[Abstract]:Sequential pattern mining is an important research field in data mining. Based on the research of classical sequential pattern mining algorithms, this paper has obtained the following research results: 1) compare and analyze the classical sequential pattern mining algorithms Apriori AllGSPSPADE and SPAM. (2) study the sequential pattern mining algorithm (Prefix Span) based on pattern growth. The theoretical analysis of the repeated projection database in the process of sequential pattern mining is given, and the upper and lower bounds of the number of projection partition in the execution of the algorithm are given. In the worst case, the general formula of generating the number of repeated projection databases. Based on the idea of prefix projection and referring to the data structure of SPAM algorithm, the two-dimensional table is used to store the location information of frequent items in the sequence database. The number of non-empty position information sets of each column in the two-dimensional table is calculated quickly and the frequency of the sequence is determined. By checking the position of the sequence prefix, the repeated mining of the same projection sub-table is avoided in the process of mining the sequence pattern.
【学位授予单位】：四川师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13

【参考文献】