UP-tree & UP-Mine: A fast method based on upper bound for frequent pattern mining from uncertain data. (November 2021)

• Record Type: Journal Article
• Authors: Davashi, Razieh
• Is Part Of: Engineering applications of artificial intelligence. Volume 106(2021)
• Publication Date: 2021-11
• Subjects: Data mining -- Frequent pattern mining -- Uncertain frequent pattern mining -- Uncertain data -- Expected support; 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.2021.104477
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: In recent years, frequent pattern mining from uncertain data has been actively researched in data mining. There are numerous exact and upper bound-based approaches for uncertain frequent pattern mining. Exact-based algorithms may produce a large data structure and need time-consuming calculations and upper bound-based algorithms may produce many false positives. As a result, these algorithms demand much time and memory. There have been efforts to resolve the problem of upper bound-based algorithms, however, all of these methods only try to tighten the upper bound of expected support for long patterns. This is while pruning infrequent short patterns has a greater impact on reducing the false positives. To overcome these drawbacks, in this paper an efficient method based on upper bound is proposed for mining uncertain frequent patterns. The proposed method uses a new T ightened u pper bound to expected support of p atterns (Tup) which has a significant effect on reducing the number of false positives by tightening the upper bound of expected support and early pruning of infrequent 2-itemsets and their supersets. Comprehensive experimental results show that the proposed method reduces memory consumption in most cases and dramatically improves the performance of exact and upper bound-based methods in terms of runtime and scalability for dense and sparse uncertain data. Highlights: A fast method based on upper bound for uncertain frequent pattern mining. It uses a new tightened upper bound to reduce the false positives, significantly. It is compared with five outstanding algorithms. It outperforms existing algorithms in terms of runtime, memory, and scalability. It can be used in a wide variety of uncertain frequent pattern mining fields.