Efficient Algorithms for Incremental Utility Mining

• Main Authors: Chih-Yang Chang, 張智揚
• Other Authors: Jieh-Shan Yeh
• Format: Others
• Language: en_US
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/00155413794077937502

碩士 === 靜宜大學 === 資訊管理學系研究所 === 95 === Traditional association rule mining strives for discovering the frequent itemsets from transaction database and mining the potential relation among the items. This can be used in many applications such as business strategic application, market strategy, retail business sell, or market basket analysis. In the real world, however, there is some information which never occurs before in some specified period, when transaction data accumulate continuously. Therefore, mining temporal association patterns from the most recent data is also important. On the other hand, utility can measure how useful or valuable an item is. The goal of utility mining is to find the high utility itemsets. Traditional association rule is a special case of utility mining, where the utility of each item is 1 and the sales quantity of each item is either 0 or 1. Utility can provide more information than association rule mining can. In addition, the valuable items found by the utility mining can help managers to make important business decisions clearly. This thesis proposed Incremental Utility Mining (IUM) and Fast Incremental Utility Mining (FIUM) Algorithms to identify the high temporal utility itemsets that users will be interested in particular time period. This thesis also extended IUM-Algorithm and FIUM-Algorithm to IUM+-Algorithm and FIUM+-Algorithm, respectively. These algorithms can efficiently discover all of the high temporal utility itemsets when the new transaction data are added in or the old transaction data are removed from the original transaction database.