Paper 6
Discovering Periodic-Correlated Patterns in Temporal DatabasesAuthors: J. N. Venkatesh, R. Uday Kiran, P. Krishna Reddy, Masaru Kitsuregawa |
AbstractThe support and periodicity are two important dimensions to determine the interestingness of a pattern in a dataset. Periodic-frequent patterns are an important class of regularities that exist in a dataset with respect to these two dimensions. Most previous models on periodic-frequent pattern mining have focused on finding all patterns in a transactional database that satisfy the user-specified minimum support (minSup) and maximum periodicity (maxPer) constraints. These models suffer from the following two obstacles: (i) Current periodic-frequent pattern models cannot handle datasets in which multiple transactions can share a common time stamp and/or transactions occur at irregular time intervals (ii) The usage of single minSup and maxPer for finding the patterns leads to the rare item problem. This paper tries to address these two obstacles by proposing a novel model to discover periodic-correlated patterns in a temporal database. Considering the input data as a temporal database addresses the first obstacle, while finding periodic-correlated patterns address the second obstacle. The proposed model employs all-confidence measure to prune the uninteresting patterns in support dimension. A new measure, called periodic-all-confidence, is being proposed to filter out uninteresting patterns in periodicity dimension. A pattern-growth algorithm has also been discussed to find periodic-correlated patterns. Experimental results show that the proposed model is efficient. Keywords: Data mining, Pattern mining, Periodic patterns, Rare item problem, Pattern-growth technique. |