Search

• Existing time series forecasting models make it difficult to extract multiple common features in the data, such as sudden irregular fluctuation trends, trend features with large fluctuation amplitudes, and factors such as long-term dependencies in the time series. To solve this problem, a heterogeneous model NeA3L that considers the common features of time series is proposed. • The proposed model realizes multiple independent parallel feature extraction modules by combining multi-task learning, which effectively captures the common features of multivariate time series the long-term dependency features in the time series and enhances the model's ability to extract common features of time series. The model realizes the feature fusion of multivariate time series by combining the attention mechanism. By designing a deep iterative LSTM network, the model handles sudden irregular fluctuation regular features. • Comparing with six state-of-the-art time series forecasting models and three traditional forecasting models on four different domain public datasets, the model outperforms nine benchmark models in all forecasting indexes, indicating that the model in this paper has the ability of universal applicability in the field of multivariate time series forecasting, which is of great significance for optimizing intelligent management and decision-making. Multivariate time series data contain a variety of common features that are difficult to extract, among which the sudden irregular fluctuation trend, the trend feature of large fluctuation amplitude, and the long-term dependency relationship in the time series have an important impact on the accuracy of the prediction model. To predict non-stationary trends in large-scale data accurately using the common characteristics of multivariate time series. A novel generalised forecasting model NeA3L based on common features of time series is developed. The NeA3L model utilizes multiple independent parallel feature extraction modules to obtain the common features of multivariate time series. It utilizes the three-layer iterative structure to deal with sudden irregular fluctuation patterns. NeA3L optimizes the network structure to realize the heterogeneous codec structure. It applies the attention mechanism between the encoder and the decoder to accomplish the multivariate prediction of the multivariate time series, which has good stability for predicting various types of multivariate data. A comparison of the NeA3L model with nine current time series prediction methods on four publicly available datasets shows that the NeA3L model outperforms the current methods in various evaluation metrics. The RMSE is improved by 2.3 %, 26.5 %, 28.9 %, and 20.84 % on average, respectively. The NeA3L model can be universally applicable to the field of multivariate time series prediction, which is important for optimizing the intelligent management and decision-making. [ABSTRACT FROM AUTHOR]

As the key step of anomaly detection for multivariate time-series (MTS) data, learning the relations among different variables has been explored by many approaches. However, most existing approaches overlook the heterogeneity among variables, that is, different types of variables (continuous numerical variables, discrete categorical variables or hybrid variables) may have different edge distributions. In this paper, we propose a novel semi-supervised anomaly detection framework based on a heterogeneous feature network (HFN) for MTS. Specifically, we first combine the embedding similarity subgraph generated by sensor embedding and the feature value similarity subgraph generated by sensor values to construct a time-series heterogeneous graph, which fully utilizes the rich heterogeneous mutual information among variables. Then, a prediction model containing nodes and channel attentions is jointly optimized to obtain better time-series representations. This approach fuses the state-of-the-art technologies of heterogeneous graph structure learning (HGSL) and representation learning. Experimental results on four sensor datasets from real-world applications demonstrate that our approach achieves more accurate anomaly detection compared to baseline methods, laying a foundation for the rapid positioning of anomalies. [ABSTRACT FROM AUTHOR]

Time-series data are widespread in real-world industrial scenarios. To recover and infer missing information in real-world applications, the problem of time-series prediction has been widely studied as a classical research topic in data mining. Deep learning architectures have been viewed as next-generation time-series prediction models. However, recent studies have shown that deep learning models are vulnerable to adversarial attacks. In this study, we prospectively examine the problem of time-series prediction adversarial attacks and propose an attack strategy for generating an adversarial time series by adding malicious perturbations to the original time series to deteriorate the performance of time-series prediction models. Specifically, a perturbation-based adversarial example generation algorithm is proposed using the gradient information of the prediction model. In practice, unlike the imperceptibility to humans in the field of image processing, time-series data are more sensitive to abnormal perturbations and there are more stringent requirements regarding the amount of perturbations. To address this challenge, we craft an adversarial time series based on the importance measurement to slightly perturb the original data. Based on comprehensive experiments conducted on real-world time-series datasets, we verify that the proposed adversarial attack methods not only effectively fool the target time-series prediction model LSTNet, they also attack state-of-the-art CNN-, RNN-, and MHANET-based models. Meanwhile, the results show that the proposed methods achieve a good transferability. That is, the adversarial examples generated for a specific prediction model can significantly affect the performance of the other methods. Moreover, through a comparison with existing adversarial attack approaches, we can see that much smaller perturbations are sufficient for the proposed importance-measurement based adversarial attack method. The methods described in this paper are significant in understanding the impact of adversarial attacks on a time-series prediction and promoting the robustness of such prediction technologies. [ABSTRACT FROM AUTHOR]

Echo state networks (ESNs) are a new approach to recurrent neural networks (RNNs) that have been successfully applied in many domains. Nevertheless, an ESN is a predictive model rather than a classifier, and methods to employ ESNs in time series classification (TSC) tasks have not yet been fully explored. In this paper, we propose a novel ESN approach named functional echo state network (FESN) for time series classification. The basic idea behind FESN is to replace the numeric variable output weights of an ESN with time-varying output-weight functions and introduce a temporal aggregation operator to the output layer that can project temporal signals into discrete class labels, thereby transforming the ESN from a predictive model into a true classifier. Subsequently, to learn the output-weight functions, a spatio-temporal aggregation learning algorithm is proposed based on orthogonal function basis expansion. By leveraging the nonlinear mapping capacity of a reservoir and the accumulation of temporal information in the time domain, FESN can not only enhance the separability of different classes in a high-dimensional functional space but can also consider the relative importance of temporal data at different time steps according to dynamic output-weight functions. Theoretical analyses and experiments on an extensive set of UCR data were conducted on FESN. The results show that FESN yields better performance than single-algorithm methods, has comparable accuracy with ensemble-based methods and exhibits acceptable computational complexity. Interestingly, for some time series datasets, we visualized some interpretable features extracted by FESN via specific patterns within the output-weight functions. [ABSTRACT FROM AUTHOR]

In this paper, we propose an unsupervised learning framework based on Genetic Programming (GP) to predict the position of any particular target event (defined by the user) in a time-series. GP is used to automatically build a library of candidate temporal features. The proposed framework receives a training set S = { ( V a ) | a = 0 ⋯ n } , where each V a is a time-series vector such that ∀ V a ∈ S , V a = { ( x t ) | t = 0 ⋯ t max } where t max is the size of the time-series. All V a ∈ S are assumed to be generated from the same environment. The proposed framework uses a divide-and-conquer strategy for the training phase. The training process of the proposed framework works as follow. The user specifies the target event that needs to be predicted (e.g., Highest value, Second Highest value , …, etc.). Then, the framework classifies the training samples into different Bins , where Bins = { ( b i ) | i = 0 ⋯ t max } , based on the time-slot t of the target event in each V a training sample. Each b i ∈ Bins will contain a subset of S . For each b i , the proposed framework further classifies its samples into statistically independent clusters. To achieve this, each b i is treated as an independent problem where GP is used to evolve programs to extract statistical features from each b i ’s members and classify them into different clusters using the K-Means algorithm. At the end of the training process, GP is used to build an ‘event detector’ that receives an unseen time-series and predicts the time-slot where the target event is expected to occur. Empirical evidence on artificially generated data and real-world data shows that the proposed framework significantly outperforms standard Radial Basis Function Networks, standard GP system, Gaussian Process regression, Linear regression, and Polynomial Regression. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, we present a method to overcome the random walk (RW) dilemma for financial time series forecasting, called swarm-based translation-invariant morphological prediction (STMP) method. It consists of a hybrid model composed of a modular morphological neural network (MMNN) combined with a particle swarm optimizer (PSO), which searches for the best time lags to optimally describe the time series phenomenon, as well as estimates the initial (sub-optimal) parameters of the MMNN (weights, architecture and number of modules). An additional optimization is performed with each particle of the PSO population (a distinct MMNN) using the back-propagation (BP) algorithm. After the MMNN parameters adjustment, we use a behavioral statistical test and a phase fix procedure to adjust time phase distortions observed in financial time series. Finally, we conduct an experimental analysis with the proposed method using four real world stock market time series, where five well-known performance metrics and a fitness function are used to assess the prediction performance. The obtained results are compared with those generated by classical models presented in the literature. [ABSTRACT FROM AUTHOR]

Abstract: Efficient processing of streaming time-series generated by remote sensors and mobile devices has become an important research area. As in traditional time-series applications, similarity matching on streaming time-series is also an essential research issue. To obtain more accurate similarity search results in many time-series applications, preprocessing is performed on the time-series before they are compared. The preprocessing removes distortions such as offset translation, amplitude scaling, linear trends, and noise inherent in time-series. In this paper, we propose an algorithm for distortion-free predictive streaming time-series matching. Similarity matching on streaming time-series is saliently different from traditional time-series in that it is not feasible to directly apply the traditional algorithms for streaming time-series. Our algorithm is distortion-free in the sense that it performs preprocessing on streaming time-series to remove offset translation and amplitude scaling distortions at the same time. Our algorithm is also predictive, since it performs streaming time-series matching against the predicted most recent subsequences in the near future, and thus improves search performance. To the best of our knowledge, no streaming time-series matching algorithm currently performs preprocessing and predicts future search results simultaneously. [Copyright &y& Elsevier]