Extracting temporal patterns from interval-based sequences

摘要: Most of the sequential patterns extraction methods proposed so far deal with composed events linked by temporal relationships based on simple precedence between instants. In many real situations, some quantitative information about event duration or inter-event delay is necessary to discriminate phenomena. We propose algorithm QTIPrefixSpan for extracting which intervals describing their position in time and are associated. It extends PrefixSpan a multi-dimensional interval clustering step representative associated patterns. Experiments simulated data show that our efficient precise even noisy contexts it improves performance former used method EM algorithm.