Mining Sequential Patterns without ‘Apriori’ Candidate Generation

L. Harada, Y. Hotta, and R. Take (Japan)


data mining, sequential pattern.


In this paper we present a new algorithm for mining sequential patterns found in a set of sequences with frequency higher than a user-specified value. In contrast to most previous step-wise algorithms that, for each processing step first generate large amounts of candidate sequential patterns ‘a priori’ from the set of frequent sequences of the previous step to later check if they are actually contained in the data, our algorithm takes advantage of some information contained in space efficient bit maps to only generate the sequential patterns that are actually found in the data and with chances to become frequent. Performance evaluation of the implementation of our algorithm shows its efficiency when applied to real-life data.

Important Links:

Go Back