A Novel Approach for Mining Similarity Profiled Temporal Association Patterns (1604.05272v1)

Abstract: The problem of frequent pattern mining from non-temporal databases is studied extensively by various researchers working in areas of data mining, temporal databases and information retrieval. However, Conventional frequent pattern algorithms are not suitable to find similar temporal association patterns from temporal databases. A Temporal database is a database which can store past, present and future information. The objective of this research is to come up with a novel approach so as to find similar temporal association patterns w.r.t user specified threshold and a given reference support time sequence using concept of Venn diagrams. For this, we maintain two types of supports called positive support and negative support values to find similar temporal association patterns of user interest. The main advantage of our method is that, it performs only a single scan of temporal database to find temporal association patterns similar to specified reference support sequence. This single database scan approach hence eliminates the huge overhead incurred when the database is scanned multiple times. The present approach also eliminates the need to compute and maintain true support values of all the subsets of temporal patterns of previous stages when computing temporal patterns of next stage.