Scalable data series subsequence matching with ULISSE

Abstract

Data series similarity search is an important operation and at the core of several analysis tasks and applications related to data series collections. Despite the fact that data series indexes enable fast similarity search, all existing indexes can only answer queries of a single length (fixed at index construction time), which is a severe limitation. In this work, we propose ULISSE, the first data series index structure designed for answering similarity search queries of variable length (within some range). Our contribution is twofold. First, we introduce a novel representation technique, which effectively and succinctly summarizes multiple sequences of different length. Based on the proposed index, we describe efficient algorithms for approximate and exact similarity search, combining disk based index visits and in-memory sequential scans. Our approach supports non Z-normalized and Z-normalized sequences, and can be used with no changes with both Euclidean Distance and Dynamic Time Warping, for answering both k-NN andrange queries. We experimentally evaluate our approach using several synthetic and real datasets. The results show that ULISSE is several times, and up to orders of magnitude more efficient in terms of both space and time cost, when compared to competing approaches. (Paper published in VLDBJ 2020) 1 Introduction Motivation. Data sequences are one of the most common data types, and they are present in almost every scientific and

Fast Data Series Indexing for In-Memory Data

“…This is true for other iSAX-based indices. For example, we could parallelize in a way similar to MESSI the ULISSE index [53], which supports queries of variable length, as well as the DPiSAX index [85], which is a distributed index operating on top of Spark (but currently not supporting parallel execution within each node of the Spark cluster). It is an interesting open problem to study whether there exist efficient parallelization techniques for indexing schemes whose tree index does not satisfy this large fanout property that would result in better perfomance than MESSI.…”

Effective and Efficient Variable-Length Data Series Analytics

“…Even though much effort has been dedicated for developping techniques for data series analytics, existing solutions for subsequence matching, motif and discord discovery are limited to fixed length queries/results. In this Ph.D. work, we propose the first scalable solutions to the variable-length version of these problems: ULISSE is the first index that supports variable-length subsequence matching over both Z-normalized and non Z-normalized sequences [15,13,14], while MAD is the first framework that implements variablelength motif and discord discovery [17,4,16].…”

“…1. ULISSE (ULtra compact Index for variable-length Similarity SEarch in data series) is the first indexing technique that supports variable-length subsequence matching for non Z-normalized and Z-normalized data series [15,13,14].…”

ProS: data series progressive k-NN similarity search and classification with probabilistic quality guarantees

“…Optimized and Approximate Similarity Search. The database community has optimized similarity search methods by using index structures [22,27,28,33,49,72,73,82,83,132,134,139,146] or fast sequential scans [112]. Recently, Echihabi et al [47,48] compared the efficiency of these methods under a unified experimental framework, showing that there is no single best method that outperforms all the rest.…”

“…In this context, progressive answers help to speed-up exact queries by stopping execution early, when it is highly probable that the current progressive answer is the exact one. Note that several data series similarity search methods support approximate query answering that can produce increasingly more accurate answers as time goes by [23,51,73,83,134,146], though, none of them provides quality guarantees on the answers. In this work, we focus on the iSAX2+ [23] and DSTree [134] methods, which exhibit superior performance at the similarity search task [47,48].…”