摘要
对于超市销售记录进行关联挖掘,项目集庞大,每次事务中涉及到项目数非常少。针对这类稀疏数据,提出了基于事务哈希表和线性对象表的FP-Tree改进算法,其只需扫描数据库一次,把相关信息压入事务哈希表和线性对象表中。当支持度和事务记录变化时,可不用重新扫描数据库或扫描数据库更新部分。试验结果验证了该改进算法相对于原算法在建树中的优势,特别在大数据集下,降低了建立FP-Tree的时间。
For the record of supermarket sales of association data mining,the project set waslarge,each transaction involved in the number of items was very small,sparse data for this and similar proposed transaction-based hash table and the linear object table of the FP-TREE improved algorithm,It scans the database only once to press into the affairs related to information and linear hash table and object table.When the change of support and transaction records,it can not re-scan the database or scan the database update part.The results are used to verify the improved algorithm related to the original FP-TREE algorithm in the achievements of the advantages,especially in large data sets,the association rule mining greatly reduces the time of construction of FP-TREE.
出处
《长江大学学报(自科版)(上旬)》
CAS
2010年第1期76-79,共4页
JOURNAL OF YANGTZE UNIVERSITY (NATURAL SCIENCE EDITION) SCI & ENG
基金
福建省自然科学基金资助项目(2009J01294)
宁德师范学院科研资助项目(2008Y007)