Discovering Top-k Periodic-Frequent Patterns in Very Large Temporal Databases

Abstract

Discovering periodic-frequent patterns in temporal databases is a challenging data mining problem with abundant applications. It involves discovering all patterns in a database that satisfy the user-specified minimum support (minSup) and maximum periodicity (maxPer) constraints. MinSup controls the minimum number of transactions in which a pattern must appear in a database. MaxPer controls the maximum time interval within which a pattern must reappear in the database. Setting an appropriate minSup and maxPer values for any given database is an open research problem. This paper addresses this open problem by proposing a solution to discover top-k periodic-frequent patterns in a temporal database. Top-k periodic-frequent patterns represent a total of k periodic-frequent patterns with the lowest periodicity value in a database. An efficient depth-first search algorithm, called Top-k Periodic-Frequent Pattern Miner (k-PFPMiner), which takes only k threshold as an input was presented to find all desired patterns in a database. Experimental results on synthetic and real-world databases demonstrate that our algorithm is memory and runtime efficient and highly scalable.