A Less Computation Approach to Reconstructing Frequent Pattern Trees in a Dynamic Database

• Main Authors: Wen-jun Chen, 陳玟君
• Other Authors: Ying-Kuei Yang
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/75096527027825418859

碩士 === 國立臺灣科技大學 === 電機工程系 === 102 === Due to rapid development of information technology that enables the enterprises to have more and more ways to get information, all industries have huge storage of their past accumulated transaction data. How to use the data mining techniques to dig out useful information hidden in large database has become an important issue. As time goes by, the number of transactions in the database will gradually increase and the amount of data will become enormously large. When transaction database has been changed, that happens often in most applications, it is necessary to perform the very time-consuming task of re-mining (meaning the re-scanning) the whole databse if traditional mining algorithm is used. The mining efficiency will be greatly improved if the task of re-scanning whole data base can be voided by reserving the information obtained at previous stage of setting up the database. Based on the concept of AFPIM (Adjusting FP-Tree Structure for Incremental Mining) algorithm, this thesis proposes a LCRFP-tree (A Less Computation Approach of Reconstructing FP-Trees) that not ony retains the frequent itemsets in original database but also records infrequent items by an IFP-tree (Infrequent Pattern-tree). The whole transaction database can therefore be incorporated in the proposed FP- and IFP-tres. Consequently, there is no need to re-scan the whole database when it is changed or updated, making the mining performance more efficient in processing time.