Sequential pattern mining with multiple minimum supports: A tree based approach

Frequent pattern mining is an important data-mining method for determining correlations among items/itemsets. Since the frequencies for various items are always varied, specifying a single minimum support cannot exactly discover interesting patterns. To solve this problem, Liu et al. propose an apriori-based method to include the concept of multiple minimum supports (MMS in short) on association rule mining. It allows user to specify MMS to reflect the different natures of items. Since the mining of sequential pattern may face the same problem, we extend the traditional definition of sequential patterns to include the concept of MMS in this study. For efficiently discovering sequential patterns with MMS, we develop a data structure, named PLMS-tree, to store all necessary information from database. After that, a pattern growth method, named MSCP-growth, is developed to discover all sequential patterns with MMS from PLMS-tree.