Similarity Search of Bounded TIDASETs within Large Time Interval Databases

Searching for similar entities within a database is a common and a daily, billions of times, performed task. Generally, similarities are calculated using common distance measures like Manhatten, Euclidian, Levenshtein, Mahalanobis or Dynamic Time Warping (DTW). In this paper, we present a similarity measure for time interval data, which allows searching for similar sets of time interval records bounded by a time window (e.g., a day, a week, or a month). We introduce three different groups of distance measures i.e., temporal order, temporal measure, and temporal relation distances. In addition, we present bitmap-based implementations for algorithms of each of the three types. We designed our solutions to perform well on large datasets and support distributed calculations. Evaluations show the out-standing performance regarding other interval related similarity measures, i.e., ARTEMIS and IBSM.