Speeding-Up the Similarity Search in Time Series Databases by Coupling Dimensionality Reduction Techniques with a Fast-and-Dirty Filter

In this paper we present a new generic frame that boosts the performance of different time series dimensionality reduction techniques by using a fast-and-dirty filter that we combine with the lower bounding condition of the dimensionality reduction technique to increase the pruning power. This fast-and-dirty filter is based on an optimal approximation of the segmented time series. The distances between these segmented time series and their approximating functions are computed and stored at indexing-time. This step is repeated using different resolution levels which correspond to different lengths of the segments. At query-time these pre-computed distances are utilized to prune those time series which are not similar to the given pattern using the least number of query-time distance computations. We conduct experiments that validate the theoretical basis of our proposed method.