Utility Pattern Mining Algorithm Based on Improved Utility Pattern Tree
Shuning Xing, Liu Fang-ai, Jiwei Wang, Lin Pang, Zhen-Guo Xu · 2015
UP-Growth is one of the most discussed high utility pattern mining algorithms based on the data structure of UP-Tree. However, the process of construction trees needs to scan database several times and spends much time in calculating. To solve these problems, an improved construction process of UP-Tree is proposed by introducing a Fast Utility Tree (FU-Tree). In this method, we introduce the LinkQueue to reduce the number of scanning the original database and adopt prefix utility to minimize the overestimated utility. The theoretical analyses and experimental results show that FU-Tree outperforms UP-Tree in the time consumption of construction trees, and enhances the efficiency of mining high utility itemsets.