E-msNSP: Efficient Negative Sequential Patterns Mining Based on Multiple Minimum Supports

Abstract

Mining negative sequential patterns (NSP) has been an important research area in data mining and knowledge discovery and it is much more challenging than mining positive sequential patterns (PSP) due to the computational complexity and search space. Only a few methods have been proposed to mine NSP and most of them only use 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. There are several methods to mine sequential patterns with multiple minimum supports (MMS), but these methods only consider PSP and do not handle NSP. So in this paper, we propose a new method, called e-msNSP, to mine NSP with multiple minimum supports. We also solve the problem of how to set up the minimum support to a sequence with negative item(s). E-msNSP consists of three major steps: (i) using the improved MS-GSP method to mine PSP with multiple minimum supports and storing all positive sequential candidates’ (PSC) related information simultaneously; (ii) using the same method in e-NSP to generate negative sequential candidates (NSC) based on above mined PSP; (iii) calculating the support of these NSC based only on the corresponding PSP and then getting NSP. To the best of our knowledge, e-msNSP is the first method to mine NSP with MMS and does not impose strict constraints. Experimental results show that the e-msNSP is highly effective and efficient.