Mining of outlier temporal patterns

Outlier temporal pattern mining problem is the study and discovery of abnormal, invalid, anomalous temporal patterns in a given temporal database. In this paper, we address the approach for mining of outlier temporal patterns with respect to a given threshold and reference. To verify if the given pattern is an outlier pattern, we compute the true support of temporal pattern and then obtain the distance between this pattern and reference temporal pattern using a novel measure. If the threshold distance computed using the proposed measure exceeds the minimum threshold limit, the pattern is treated as an outlier. Discovery and prediction of repeating temporal patterns and understanding the behavior of temporal pattern trends is quite challenging in the case of time stamped temporal datasets. At present, existing algorithms for temporal pattern mining do not have methods to reveal pattern which are emerging, seasonal and diminishing. Determining similar temporal patterns and unearthing eccentric patterns require an efficient dissimilarity measure. This research addresses the similarity measure for revealing outlier temporal patterns.