Mining spatiotemporal association rule based on prevalent sequential patterns

Spatiotemporal association rules mining algorithms can effectively mine the association relationships between spatiotemporal event patterns, which play an important role in pollution control, weather forecasting, and other fields. However, most existing association rule mining algorithms require quantitative time intervals between predefined patterns, and these algorithms do not consider sequential patterns between different spaces. To address these issues, we propose an efficient algorithm for mining spatiotemporal association rules based on prevalent sequential patterns (STARPSP). Firstly, a cross-spatiotemporal sequence dataset is constructed by discretizing sequence data from different spaces to mine both sequential patterns in the same space and between various spaces. Secondly, the PrefixSpan algorithm is extended to mine spatiotemporal prevalent sequential patterns. Thirdly, spatiotemporal association rules are automatically obtained based on prevalent sequential patterns with time intervals. Finally, STARPSP is compared with other spatiotemporal sequential patterns and association rules mining algorithms on three real datasets. Our experiments show that STARPSP outperforms the other algorithms with respect to both execution time and the number of generated rules. Furthermore, the rules generated by STARPSP contain more spatiotemporal information, which is critical in decision-making.