REDEFINING TRANSITIONAL PATTERNS FOR VALID ON DYNAMIC DATABASES

Abstract

The researchers have focused on extracting time included knowledge that reveals the behavior of item sets, such
as finding the patterns that are more present on a specific time period; finding the specific time points, where the
frequency of an item set before or after a time point increases or decreases significantly, etc. Many previous works also
consider the time points in Frequent Pattern Mining (FPM) studies and presented temporal mining algorithms (TPM).
All these TPM based works included time as an element, but nowhere in these works how dynamically is a frequent
pattern P changing its behavior in the database D. The authors Wan and An presented “Transitional Patterns” which
represent item sets whose support significantly changes from one time point to another time point in the database.
Primarily Wan and a focused on finding time points at which negative (or positive) transitional patterns decreases (or
increases) their support significantly with the change of time. TP-Mine algorithm has the limitation of when we add new
transactions to the database and by reapplying the TP-Mine algorithm on the updated database, already identified time
points may not valid on the updated database. The motive of this paper is to address the limitation existing in the TPMine algorithm proposed by Wan and An and it is done by redefining the definition of transitional pattern and proposing
an efficient two scan algorithm for finding redefined transitional patterns in parallel to the extraction of frequent
patterns.