An efficient sliding window algorithm for detection of sequential patterns

Recently a growing number of applications monitor the physical world by tracking sensor data and detecting values, trends or patterns of interest. We focus on the problem of detecting sequential patterns with complex predicates over sensor data, and present an algorithm that efficiently pre-computes which pattern predicates' checks can be skipped at query compile-time, so that the processing window can slide with only necessary checks being actually performed against the sensor data at run-time. Implementation and evaluation of the proposed approach confirms its efficiency when compared to previously proposed approaches.