Fast and Accurate Time Series Classification with WEASEL

Abstract

Time series (TS) occur in many scientific and commercial applications, ranging from earth surveillance to industry automation to the smart grids. An important type of TS analysis is classification, which can, for instance, improve energy load forecasting in smart grids by detecting the types of electronic devices based on their energy consumption profiles recorded by automatic sensors. Such sensor-driven applications are very often characterized by (a) very long TS and (b) very large TS datasets needing classification. However, current methods to time series classification (TSC) cannot cope with such data volumes at acceptable accuracy; they are either scalable but offer only inferior classification quality, or they achieve state-of-the-art classification quality but cannot scale to large data volumes. In this paper, we present WEASEL (Word ExtrAction for time SEries cLassification), a novel TSC method which is both scalable and accurate. Like other state-of-the-art TSC methods, WEASEL transforms time series into feature vectors, using a sliding-window approach, which are then analyzed through a machine learning classifier. The novelty of WEASEL lies in its specific method for deriving features, resulting in a much smaller yet much more discriminative feature set. On the popular UCR benchmark of 85 TS datasets, WEASEL is more accurate than the best current non-ensemble algorithms at orders-of-magnitude lower classification and training times, and it is almost as accurate as ensemble classifiers, whose computational complexity makes them inapplicable even for mid-size datasets. The outstanding robustness of WEASEL is also confirmed by experiments on two real smart grid datasets, where it out-of-the-box achieves almost the same accuracy as highly tuned, domain-specific methods

Scalable Dictionary Classifiers for Time Series Classification

“…Hence, more resources are spent in training to find the most discriminative features for use in a single classifier. While prediction time is not covered in our results, it should be noted that in its results WEASEL is superior to BOSS and other ensemble based techniques in this respect [21]. WEASEL performs a parameter search for mean and a reduced range of l and uses a 10-fold cross-validation to determine the performance of each set.…”

“…WEASEL [21] is a dictionary based classifier that is an extension of BOSS. WEASEL is a single classifier rather than an ensemble.…”

“…Thirdly, the histograms can be very large, so storing them all for each base classifier also requires a significant memory commitment for large problems. One proposed method to solve this, the BOSS vector space (BOSS-VS) classifier [20], has been shown to be significantly less accurate than the full BOSS [21,18]. We investigate whether we can mitigate against these problems without this significant loss in accuracy.…”

“…We investigate whether we can mitigate against these problems without this significant loss in accuracy. Patrick Schäfer, the author of the BOSS paper, recently proposed an alternative dictionary based classifier called WEASEL [21]. WEASEL is claimed to be a more accurate approach than BOSS, not more efficient.…”

A machine learning approach for the spatiotemporal forecasting of ecological phenomena using dates of species occurrence records

“…Among several applications, the time-series classification has been used to detect 'normal' or 'abnormal' heart rhythms in ElectroCardioGrams (Kampouraki, Manis & Nikou, 2009) and to identify insect species from the frequencies of their wing-beats (Potamitis, Rigakis, & Fysarakis 2015). The classification can be done using the 'raw' time series or a set of predictors which summarise their properties (i.e., the 'features', in machine learning parlance) (Schäfer & Leser, 2017). The 'raw' series approaches calculate a point-bypoint similarity with the time series of known classes.…”