HUNSPM: An efficient algorithm for mining high utility negative sequential patterns

Abstract

Recently, High Utility Sequential Patterns (HUSP) Mining plays more and more important role in many applications. However, the current HUSP mining algorithms only considers Positive Sequential Patterns (PSP), therefore limiting their ability to hand Negative Sequential Patterns (NSP). In some applications, NSP may bring more valuable information than PSP due to the absent elements. Although many algorithms have been proposed to mine NSP with frequency-based model, they are not suitable for mining NSP with utility-based model because an item in a NSP may have multiple utilities. In this paper, we introduce a new algorithm HUNSPM to identify high utility negative sequential patterns (HUNSP). HUNSPM consists of four major steps: (i) using traditional HUSP mining algorithms to mine all HUPSP; (ii) generating High Utility Negative Sequential Candidates (HUNSC) based on above mined HUPSP and storing all related information about HUNSC in the PNU-List; (iii) removing unpromising HUNSC and calculating the utility of promising HUNSC; (iv) getting all HUNSP. We have tired our best to indicate that HUNSPM is the first method to find NSP on the basis of minimum utility. The consequence of the experiment propose that the HUNSPM is efficient and highly effective.