期刊名称:International Journal of Computer Science and Information Technologies
电子版ISSN:0975-9646
出版年度:2016
卷号:7
期号:3
页码:1607-1614
出版社:TechScience Publications
摘要:Pattern mining is a data mining method thatinvolves finding existing patterns in data. The main objectiveis to uncover the hidden regularities in the data. Until nowmost of the research in pattern mining has been exclusivelyfocused on frequent item sets, however in some situation it isinteresting to find patterns that are rare instead of frequent orpatterns that reflect a negative co relation between items.These patterns are referred to as rare and negative patterns.For example in jewelery sales data, sales of diamond watchesare rare however patterns involving the selling of diamondwatches could be interesting. Rare items occur rarely but areof special interest as they may highlight exceptional behaviorin the data which is likely of interest. This paper propose amethod that uses a bidirectional approach in which we arecombining both bottom up and top down method to generatenon zero rare item set. The bottom-up search starts from 1-itemset and proceeds upto n-itemsets as in Apriori while thetop-down search starts from n item sets and proceeds down to1-itemset.The main idea is to use the upward closure propertyin the bottom-up approach and downward closure property inthe top-down search to reduce the number of candidatesgenerated as well as the number of scans over the dataset.
关键词:Rare item set; Non zero rare item set; Minimal;Rare item set; Rare Association Rule.