Searching Maximal Frequent Itemsets using Random Two-Way Pruning

• Main Author: 楊哲綜
• Other Authors: 黃仁鵬
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/90895980239900776702

碩士 === 南台科技大學 === 資訊管理系 === 92 === We propose two novel approaches for searching maximal frequent itemsets. They applied the concepts of genetic algorithm and random pruning on searching maximal frequent itemsets from transaction databases. The natures of these two algorithms make it possible to locate the frequent itemsets on their searching trees directly and free from enumerating all the frequent level by level. The first one is called by GAMax (Genetic Algorithm Based Searching of Maximal Frequent Itemsets). GAMax based on the concepts of finding optimal solutions to find the maximal frequent itemsets in transaction database. It is an alternative for solving pattern generation in association rules mining. The objective is to identify maximal frequent itemsets in lexicographic tree, needless to enumerate all the frequent itemsets level by level. Through well-defined fitness functions and operators, GAMax manipulates the specific meaningful codes in search space iteratively to determine the itemsets. This research transits the problems of generate the maximal frequent itemsets of association rules mining into optimization problems and implement using genetic algorithm. The other is called by RP_MFI (Searching Maximal Frequent Itemsets with Random Pruning). It was developed through improving the performance of GAMax. To accelerate the convergence of chromosomes and avoid the duplicate results of convergence are the points of improvements. RP_MFI applied the random pruning and two-way pruning to achieve better performance than GAMax. The experimental results of RP_MFI prove its efficiency. It is the new concepts to searching randomly and also expected to open new visions in combining genetic algorithm and data mining. They are both the flourishing areas of research and promising in current.