Efficient fuzzy-based high utility pattern computing and analyzing approach with temporal properties

Abstract

Fuzzy logic in soft computing deals with intuitive and comprehensive intelligence to find solutions to problems in the uncertain real world. Considering the fuzzy set concept and knowledge discovery of utility-driven patterns simultaneously, quantities of sets of items hidden within vast data can be represented in an easy-to-understand linguistic representation. This can lead to more reasonable decision-making. Together with these fascinating results, temporal fuzzy utility pattern analysis has emerged as a significant area in the last few years to consider the duration of transactions in temporal quantitative data. The latest temporal approaches have improved resource efficiency by storing information on patterns with efficient data structures. However, although a list-based approach is known to be robust and follows a mechanism that does not generate candidates, it requires explosive comparison operations that are unsuitable for processing long-length patterns, especially in big data analysis. To solve this issue, we present a novel indexed list-based structure along with a data analysis method designed to allow rapid pattern growth as well as prevent the generation of candidates for discovering high temporal fuzzy utility patterns. Performance tests on real and synthetic datasets demonstrate that the proposed approach exhibits superior time efficiency and scalability relative to state-of-the-art methods with minimal compromise in memory, all while extracting accurate results. Moreover, comprehensive experiments demonstrate the capability of the proposed method for practical use cases and its effectiveness in search space pruning.