Efficient automatic discovery of ‘hot’ itemsets

• Published in: Information processing letters Vol. 90; no. 2; pp. 65 - 72
• Main Authors: Kouris, Ioannis N., Makris, Christos H., Tsakalidis, Athanasios K.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 30.04.2004; Elsevier Science; Elsevier Sequoia S.A
• Subjects: Algorithmics. Computability. Computer arithmetics; Algorithms; Applied sciences; Association rules; Computer science; control theory; systems; Data mining; Exact sciences and technology; Experiments; Hot itemsets; Information systems. Data bases; Mathematical analysis; Memory organisation. Data processing; Software; Studies; Theoretical computing; Databases; Algorithms; Data mining; Association rules; Hot itemsets; Database; Computer theory; Itemset; Information processing
• ISSN: 0020-0190; 1872-6119
• DOI: 10.1016/j.ipl.2004.01.013

In real life applications the dominant model of the single support, which assumed all itemsets to be of the same nature and importance proved defective. The non-homogeneity of the itemsets on one hand and the non-uniformity of their number of appearances on the other require that we use different approaches. Some techniques have been proposed thus far trying to address these inefficiencies, but then new more demanding questions arose such as which itemsets are more interesting than others, what distinguishes them and how should they be identified, and finally how they should be effectively handled. Furthermore one common drawback of all approaches is that they have a tremendous lag in discovering new relationships and work only with long existing relationships or patterns. We propose a method that finds what we define as ‘hot’ itemsets in our database, deals with all problems described above and yet proves very efficient.