Uncertain Time Series Classification

Time series analysis has gained a lot of interest during the last decade with diverse applications in a large range of domains such as medicine, physic, and industry. The field of time series classification has been particularly active recently with the development of more and more efficient methods. However, the existing methods assume that the input time series is free of uncertainty. However, there are applications in which uncertainty is so important that it can not be neglected. This project aims to build efficient, robust, and interpretable classification methods for uncertain time series.

Flood forecasting in small catchments using deep learning LSTM networks

<p>Long short-term memory (LSTM) networks are able to learn and replicate the relationships of multiple climate and hydrological temporal variables, and therefore are theoretically suitable for data driven modelling and forecasting of rainfall-runoff behavior. However, they inherit some prediction errors occasionally found in data-driven models: phase shift errors, oscillations and total failures. The phase shift error is a particularly significant challenge due to its occurrence when using hourly precipitation and runoff data for catchments with short response times.</p><p>In order to detect and eliminate these errors, we investigated four approaches, of which the first two are of structural nature, while the last two modify the input time series by certain transformations: <br>1. The use of encoder-decoder architectures for LSTM networks. <br>2. Offsetting the start of the flood forecast to the forecast time step of interest. <br>3. The inversion of the input time series. <br>4. Including subsequently observed precipitation data as a “best precipitation forecast”.</p><p>We tested the four approaches on five different pilot catchments located in Saxony, Germany with relatively short response times. The results show no advantage of the structural approaches. In contrast, the modification of the input time series shows potential for improving the predictive quality of flood forecasting in a potential operational application.</p>

GENDIS: Genetic Discovery of Shapelets

In the time series classification domain, shapelets are subsequences that are discriminative of a certain class. It has been shown that classifiers are able to achieve state-of-the-art results by taking the distances from the input time series to different discriminative shapelets as the input. Additionally, these shapelets can be visualized and thus possess an interpretable characteristic, making them appealing in critical domains, where longitudinal data are ubiquitous. In this study, a new paradigm for shapelet discovery is proposed, which is based on evolutionary computation. The advantages of the proposed approach are that: (i) it is gradient-free, which could allow escaping from local optima more easily and supports non-differentiable objectives; (ii) no brute-force search is required, making the algorithm scalable; (iii) the total amount of shapelets and the length of each of these shapelets are evolved jointly with the shapelets themselves, alleviating the need to specify this beforehand; (iv) entire sets are evaluated at once as opposed to single shapelets, which results in smaller final sets with fewer similar shapelets that result in similar predictive performances; and (v) the discovered shapelets do not need to be a subsequence of the input time series. We present the results of the experiments, which validate the enumerated advantages.

Detecting dynamical anomalies in time series from different palaeoclimate proxy archives using windowed recurrence network analysis

Analysing palaeoclimate proxy time series using windowed recurrence network analysis (wRNA) has been shown to provide valuable information on past climate variability. In turn, it has also been found that the robustness of the obtained results differs among proxies from different palaeoclimate archives. To systematically test the suitability of wRNA for studying different types of palaeoclimate proxy time series, we use the framework of forward proxy modelling. For this, we create artificial input time series with different properties and compare the areawise significant anomalies detected using wRNA of the input and the model output time series. Also, taking into account results for general filtering of different time series, we find that the variability of the network transitivity is altered for stochastic input time series while being rather robust for deterministic input. In terms of significant anomalies of the network transitivity, we observe that these anomalies may be missed by proxies from tree and lake archives after the non-linear filtering by the corresponding proxy system models. For proxies from speleothems, we additionally observe falsely identified significant anomalies that are not present in the input time series. Finally, for proxies from ice cores, the wRNA results show the best correspondence to those for the input data. Our results contribute to improve the interpretation of windowed recurrence network analysis results obtained from real-world palaeoclimate time series.

Detecting dynamical anomalies in time series from different palaeoclimate proxy archives using windowed recurrence network analysis

<p>Analysing palaeoclimate proxy time series using windowed recurrence network analysis (wRNA) has been shown to provide valuable information on past climate variability. In turn, it has also been found that the robustness of the obtained results differs among proxies from different palaeoclimate archives. To systematically test the suitability of wRNA for studying different types of palaeoclimate proxy time series, we use the framework of forward proxy modelling. For this, we create artificial input time series with different properties and compare the areawise significant anomalies detected using wRNA of the input and the model output time series. Also, taking into account results for general filtering of different time series, we find that the variability of the network transitivity is altered for stochastic input time series while being rather robust for deterministic input. In terms of significant anomalies of the network transitivity, we observe that these anomalies may be missed by proxies from tree and lake archives after the non-linear filtering by the corresponding proxy system models. For proxies from speleothems, we additionally observe falsely identified significant anomalies that are not present in the input time series. Finally, for proxies from ice cores, the wRNA results show the best correspondence with those for the input data. Our results contribute to improve the interpretation of windowed recurrence network analysis results obtained from real-world palaeoclimate time series.</p>

Detecting dynamical anomalies in time series from different palaeoclimate proxy archives using windowed recurrence network analysis

Analysing palaeoclimate proxy time series using windowed recurrence network analysis (wRNA) has been shown to provide valuable information on past climate variability. In turn, it has also been found that the robustness of the obtained results differs among proxies from different palaeoclimate archives. To systematically test the suitability of wRNA for studying different types of palaeoclimate proxy time series, we use the framework of forward proxy modelling. For this, we create artificial input time series with different properties and, in a first step, compare the time series properties of the input and the model output time series. In a second step, we compare the areawise significant anomalies detected using wRNA. For proxies from tree and lake archives, we find that significant anomalies present in the input time series are sometimes missed in the input time series after the nonlinear filtering by the corresponding models. For proxies from speleothems, we observe falsely identified significant anomalies that are not present in the input time series. Finally, for proxies from ice cores, the wRNA results show the best correspondence with those for the input data. Our results contribute to improve the interpretation of windowed recurrence network analysis results obtained from real-world palaeoclimate time series.