• 论文 •    

基于全局图遍历的加权频繁模式挖掘算法

耿汝年,董祥军,须文波   

  1. 1.江南大学 信息工程学院,江苏无锡214122;2.山东轻工业学院 信息科学与技术学院, 山东济南250353
  • 出版日期:2008-06-15 发布日期:2008-06-25

Weighted frequent patterns mining algorithm based on global graph traversals

GENG Ru-nian, DONG Xiang-jun, XU Wen-bo   

  1. 1.School of Information Technology, Jiangnan University, Wuxi 214122, China; 2.School of Information Science & Technology, Shandong Institute of Light Industry, Jinan 250353, China
  • Online:2008-06-15 Published:2008-06-25

摘要: 为解决加权遍历模式挖掘问题,提出了一种从边加权有向图到顶点加权有向图的变换模型。基于该模型,提出了基于全局图遍历加权频繁模式挖掘算法。在该算法中,利用图全局拓扑信息评估遍历模式的加权支持度,将剪枝问题转化为模式可扩展性问题,进而利用可扩展模式的向下闭合特性产生候选模式集。实验结果表明,该算法是一个高效的基于图遍历的加权频繁模式挖掘算法。

关键词: 数据挖掘, 加权有向图, 遍历模式, 频繁模式, 支持度界

Abstract: A transformation model from Edge-Weighted Directed Graph (EWDG) to Vertex-Weighted Directed Graph (VWDG) was proposed to solve the problem of weighted traversal pattern mining. Based on this model, a new algorithm called Global Graph Traversals-based Weighted Frequent Patterns Miner (GGTWFPMiner) was presented. This algorithm evaluated traversal patterns' weighted support by the global topology of graph, and converted patterns' pruning problem into corresponding patterns' scalability problem, and then generated candidate sets of patterns by using the downward closure property between scalable patterns. Experimental results showed that GGTWFPMiner was an effective weighted frequent patterns mining algorithm based on graph traversals.

Key words: data mining, weighted directed graph, traversal patterns, frequent patterns, support bound

中图分类号: