Exact pattern matching is a method of localizing arbitrarily sized patterns in time series data. To date, the problem of exact pattern matching has only been fully addressed for one query pattern on one time series with a single best match location.
This paper addresses the broader problem of finding the top-K pattern matches for a multidimensional time series pattern in a large multidimensional time series. The problem is addressed in two stages using an algorithm that combines ideas from the fields of data mining and bi-clustering. The first stage of the algorithm addresses selecting the dimension subset that matches the query pattern and locating the matching pattern. The second stage of the algorithm addresses the problem of finding the top-K matches of the pattern in the selected time series dimensions.
The performance of the proposed algorithm is evaluated against the best single dimensional exact pattern matching algorithm on real and simulated data.