A similarity measure for outlier detection in timestamped temporal databases

Outlier Detection is one of the important research problems in temporal data mining. A pattern in time stamped temporal database is a sequence of probability values. Finding outliers from time stamped temporal databases requires suitable dissimilarity measure to find dissimilarity between input pattern and reference pattern which is of user interest. In the current work, the objective is to propose a dissimilarity measure for detecting outliers. We also come up with expression for deviation which is used to compute dissimilarity degree using proposed measure. The dissimilarity measure has lower bound value (0) and upper bound (1) respectively.