iSAX: disk-aware mining and indexing of massive time series datasets

Abstract

Current research in indexing and mining time series data has produced many interesting algorithms and representations. However, the algorithms and the size of data considered have generally not been representative of the increasingly massive datasets encountered in science, engineering, and business domains. In this work, we introduce a novel multi-resolution symbolic representation which can be used to index datasets which are several orders of magnitude larger than anything else considered in the literature. To demonstrate the utility of this representation, we constructed a simple tree-based index structure which facilitates fast exact search and orders of magnitude faster, approximate search. For example, with a database of one-hundred million time series, the approximate search can retrieve high quality nearest neighbors in slightly over a second, whereas a sequential scan would take tens of minutes. Our experimental evaluation demonstrates that our representation allows index performance to scale well with increasing dataset sizes. Additionally, we provide analysis concerning parameter sensitivity, approximate search effectiveness, and lower bound comparisons between time series representations in a bit constrained environment. We further show how to exploit the combination of both exact and approximate search as sub-routines in data mining algorithms, allowing for the exact mining of truly massive real world datasets, containing tens of millions of time series.Responsible editor: Bart Goethals.

Finding Time Series Motifs in Disk-Resident Data

“…star light curve catalogs), we plan to consider parallelization, given that the search for different group sizes can easily be delegated to different processors. Because it has recently been shown that DNA can be meaningfully transformed into time series [23], we plan to use DAME to Explore DNA (DAME EDNA), in particular, the DNA of Phalangeroidea (possums). Another avenue of research is to modify the algorithm to find multidimensional motifs in databases of similar scale.…”

DPiSAX: Massively Distributed Partitioned iSAX

“…There are several techniques, providing lower bounds by segmenting time series. Here, we use a popular method, called indexable Symbolic Aggregate approXimation (iSAX) representation [15], [16]. The iSAX representation will be used to represent time series in our index.…”

“…Note that previous studies have shown that the iSAX index is robust with respect to the choice of parameters (word length, cardinality, leaf threshold) [16], [5], [21]. Moreover, it can also be used to answer queries with the Dynamic Time Warping (DTW) distance, through the use of the corresponding lower bounding envelope [9].…”

“…In order to improve the performance of such similarity queries, indexing is one of the most popular techniques [6], which has been successfully used in a variety of settings and applications [7], [16], [3], [17], [5], [20]. Although recent studies have shown that in certain cases sequential scans can be very efficient [14], [18], such techniques are only advantageous when the database consists of a single, long time series, and query answers are small subsequences of this long time series.…”

Particle Swarm Optimization of Information-Content Weighting of Symbolic Aggregate Approximation

“…There are other versions and extensions of SAX [11], [25], [26], [28]. These versions use it for other applications or apply it to index massive datasets, or compute MINDIST differently [18].…”