Mining Frequent Tree-like Patterns in Large Datasets

• Main Author: 許士俊
• Other Authors: 陳宗禧
• Format: Others
• Language: en_US
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/12047818237781806568

碩士 === 長榮大學 === 經營管理研究所 === 92 === Frequent sequential pattern mining is an important domain for data mining. In this thesis, we present a new data mining scheme to explore the hierarchical structure like tree represented the relationship of each item of sequences, named tree-like patterns. By tree-like patterns, we clear to find out the relation of items between the cause and effect. For counting support value, we propose a scheme which counts support of tree-like patterns by queue structure and efficient count the support values. We also present an efficient scheme to count the frequency of tree-like patterns on a sequence by dynamic programming. We could understand the importance of tree-like patterns on a sequence. In addition, we present two formulas to compute the significance of sequences, which describes the couple degree of items and tree-like patterns on a sequence. The higher value of significance means tree-like patterns with tighter couple on a sequence. Final, we compare characters of different patterns with ours. We have more characters and applications widely.