A geometric approach for mining sequential patterns in interval-based data streams

摘要: Almost all activities observed in nowadays applications are correlated with a timing sequence. Users mainly looking for interesting sequences out of such data. Sequential pattern mining algorithms aim at finding frequent sequences. Usually, the mined have durations that represent time intervals between their starting and ending points. The majority sequential approaches dealt as single point event thus lost valuable information collected patterns. Recently, some carefully considered this interval-based nature events, but they major limitations. They concentrate only on order events without taking gaps them into account usually employ binary representation to describe To resolve these problems, we propose PIVOTMiner, an data algorithm using geometric approach intervals. Noisy can be served fuzzy set retrieved from PIVOTMiner flexibly work presented any number not necessarily aligned interval particular utilize sequence stream need create samples. Our experimental results both synthetic real-world smart home datasets show our patterns richer than those most state-of-the-art while spending considerably smaller running times.