Mining non-redundant closed flexible periodic patterns. (March 2018)

• Record Type: Journal Article
• Authors: Akther, Sayma; Rezaul Karim, Md.; Samiullah, Md.; Ahmed, Chowdhury Farhan
• Is Part Of: Engineering applications of artificial intelligence. Volume 69(2017:Sep.)
• Page Start: 1
• Page End: 23
• Publication Date: 2018-03
• Subjects: Data mining -- Time series databases -- Periodic patterns -- Closed periodic patterns -- Flexible patterns -- Non-redundant patterns; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2017.11.005
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Mining periodic patterns from time series databases is needed to predict any future situation. Flexible pattern mining is a special kind of periodic pattern mining where intermediate events can be overlooked purposely. Mining such patterns from time series data is advantageous due to its capability of modeling various real life scenarios. The goal of mining closed flexible patterns is to avoid unnecessary flexible patterns but preserving the same information of a complete set of patterns. Though it has wide range of application domains, existing algorithms failed to mine closed flexible patterns without generating any false positive, i.e. non-closed and ∕ or redundant patterns. In this paper, a new algorithm N R C F P (Non-Redundant Closed Flexible Pattern) has been proposed that generates complete set of non-redundant closed flexible patterns in time series databases. Three pruning techniques- B a c k S c a n (existing), R a n g e S c a n (proposed) and C o l u m n - p r u n i n g (proposed) have been applied to avoid generation of non-closed patterns, redundant flexible patterns and fictitious patterns. Proposed N R C F P efficiently mines non-redundant closed flexible periodic patterns. The performance of our algorithm has been extensively analyzed using several real-life databases based on runtime and memory consumption and compared with existing state-of-the-arts approach to prove effectiveness of the algorithm with respect to required processing time and memory consumption. Some applications of our proposed algorithm in various real life domains are discussed. Highlights: Designed new algorithm to mine non-redundant closed flexible periodic pattern. Proposed RangeScan technique to exhaust non-closed and redundant flexible pattern. Devised ColumnPruning technique to optimize index table to decoy fictitious pattern. Effective use of existing BackScan technique to prune non-closed flexible patterns. Extensive performance analyses have been performed using real life datasets.