FastTIRP: Efficient Discovery of Time-Interval Related Patterns.

Finding frequent patterns in discrete sequences of symbols or events can be useful to understand data, support decision-making and make predictions. However, many studies on analyzing event sequence do not consider the duration of events, and thus the complex time relationships between them (e.g. an event may start at the same time as another event but end before). To find frequent sequential patterns in data where events have a start and end time, an emerging topic is time-interval related pattern (TIRP) mining. Several algorithms have been proposed for this task but efficiency remains a major issue due to the very large search space. To provide a more efficient algorithm for TIRP mining, this paper presents a novel algorithm called FastTIRP. It utilizes a novel Pair Support Pruning (PSP) optimization to reduce the search space. Experiments show that FastTIRP outperforms the state-of-the-art VertTIRP algorithm in terms of runtime on four benchmark datasets.