E-msNFIS: An Efficient Method for Mining Negative Frequent Itemsets based on Multiple Minimum Supports

Abstract

Negative frequent itemsets (NFIS), which refer to frequent itemsets with non-occurring and occurring item(s) like (a 1 a 2 ¬a 3 a 4), have become increasingly important in real applications, such as bioinformatics and healthcare management. Very few methods have been proposed to mine NFIS and most of them only satisfy the user-specified single minimum support, which implicitly assumes that all items in the database are of the same nature or of similar frequencies in the database. This is often not the case in real-life applications. Although several methods have been proposed to mine frequent itemsets with multiple minimum supports (MMS), these methods only mine positive frequent itemsets (PFIS) and do not handle NFIS. So in this paper, we propose an efficient method, called e-msNFIS, to efficiently identify NSP with MMS by only using the identified PFIS without re-scanning database. We also solve the problem of how to set up minimum support to an itemset with negative item(s). To the best of our knowledge, e-msNFIS is the first method to mine NFIS with MMS. Experimental results on real datasets show that the e-msNFIS is highly effective and efficient.