Mining Fuzzy Common Sequential Rules with Fuzzy Time-Interval in Quantitative Sequence Databases

Abstract

There are two kinds of sequential rules. They are classical sequential rules and common sequential rules. The common sequential rules present the relationship between unordered itemsets in which all the items in the antecedent part have to appear before the ones in the consequent part. All existing algorithms for mining common sequential rules can not apply to quantitative sequence databases. Furthermore, the common sequential rules found so far did not yet reveal the time gap about the apperance of itemsets in its antecedent and consequent parts. The purpose of this article is to overcome the two disadvantages mentioned above. Specifically, the article proposes an algorithm called IFERMiner to discover common sequential rules in quantitative sequence databases, where the time gap about appearance of two attribute sets in its antecedent and consequent parts is taken account. This algorithm was developed from the ERMiner algorithm that is the most efficient algorithm to discover common sequential rules in transactional sequence databases now. The computational complexity of the IFERMiner algorithm is also shown in the article and it is polynomial. The FCSI rules found out by the IFERMiner algorithm are useful for marketing domain and market analysis.