Maximal Mixed-Drove Co-occurrence Patterns
[ 1 ] Instytut Informatyki, Wydział Informatyki i Telekomunikacji, Politechnika Poznańska | [ P ] pracownik
2023
artykuł naukowy
angielski
- Spatial data mining
- Co-location patterns
- Spatio-temporal data
- MDCOPs
EN One of the interesting types of spatio-temporal patterns is the Mixed-Drove Co-occurrence Pattern (MDCOP), which represents types of objects frequently located together in both space and time. To discover such patterns, methods based on the well-known Apriori strategy are widely used. They involve determining multi-element MDCOPs by building them up iteratively starting with the two-element patterns and then successively adding another element in each iteration. This approach can be very costly, particularly when the data is dense enough to form patterns of significant size. In this paper, we introduce a definition of a new pattern type called a Maximal Mixed-Drove Co-occurrence Pattern. We also propose a new algorithm MAXMDCOP-Miner, which resigns from popular Apriori strategy of generating candidates and, therefore, can discover long pattern without processing all their subsets. Experiments performed on synthetic and real datasets show that MAXMDCOP-Miner has high performance, in particular for dense datasets or tasks with low user-defined thresholds of spatial or time prevalence.
20.10.2022
2005 - 2028
CC BY (uznanie autorstwa)
czasopismo hybrydowe
ostateczna wersja opublikowana
przed opublikowaniem
100
5,9 [Lista 2022]