Fuzzy Utility Mining on Temporal Data

Compared with traditional high-utility itemset mining (HUIM) algorithms, the fuzzy utility mining (FUM) algorithm not only considers the quantity and the unit utility of an itemset but also provides a comprehensible way of transforming quantity into several semantic terms. However, in real-life applications, items often have time attributes, such as seasonal vegetables, weather information, and accidents, for which the HUIM and FUM algorithms do not work very well. Especially in large cities, traffic congestion always occurs during the peak time. This case will seriously lower the roads' usage. Hence, researchers proposed the temporal-based fuzzy utility itemset mining (TFUIM) technology to analyze these temporal data. In this paper, we first develop a novel list-based TFUIM algorithm, namely FUMT. The newly designed temporal fuzzy-list structure compresses all key information about temporal fuzzy itemsets to reduce the number of database scans. In addition, according to the list structure, FUMT is also a one-phase algorithm, which avoids the generate-and-test paradigm. Finally, extensive experiments show that the novel algorithm performs better than the state-of-the-art algorithms in terms of runtime and memory consumption on sparse and dense datasets.