Time Series Classification Approach Research Articles

MTS2Graph: Interpretable multivariate time series classification with temporal evolving graphs

Conventional time series classification approaches based on bags of patterns or shapelets face significant challenges in dealing with a vast amount of feature candidates from high-dimensional multivariate data. In contrast, deep neural networks can learn low-dimensional features efficiently, and in particular, convolutional neural networks have shown promising results in classifying multivariate time series data. A key factor in the success of deep neural networks is this astonishing expressive power. However, this power comes at the cost of complex, black-boxed models, conflicting with the goals of building reliable and human-understandable models. In this work11The code can be found at https://github.com/raneeny/MTS2Graph., we introduce a new interpretable framework for multivariate time series data that by extracting and clustering the input quantifies the contribution of time-varying input variables and each signal’s role to the classification. We construct a graph that captures the temporal relationship between the extracted patterns for each layer and propose an effective merging strategy to aggregate those graphs into one. Finally, a graph embedding algorithm generates new representations of the created interpretable time-series features. Our extensive experiments indicate the benefit of our time-aware graph-based representation in multivariate time series classification while enriching them with more interpretability.

Interpretable synthetic signals for explainable one-class time-series classification

This research paper introduces an innovative approach for explainable one-class time-series classification (XOCTSC). The proposed method involves generating pseudounseen synthetic signals by altering the amplitude and cycle of the original signals. Subsequently, a classification process is performed to distinguish between the original and synthetic signals, and the resulting model is applied to testing data. Instances classified as synthetic classes are treated as unseen classes, and the dissimilarity with the training data can be elucidated through an explanation of the synthetic class creation process. This approach aims to enhance the interpretability of one-class time-series classification models by providing insights into the reasoning behind their decisions. The proposed method is demonstrated with a ballistocardiogram (BCG) signal for the breathing dataset and an electroencephalogram (EEG) signal for the epilepsy dataset. The proposed method recognizes BCG amplitude reduction during breath holding. Moreover, EEG cycle changes during epileptic seizures are observed across multiple channels. These observations align with actual epilepsy symptoms and breathing behaviors.

A New Time-Series Classification Approach for Human Activity Recognition with Data Augmentation

A New Time-Series Classification Approach for Human Activity Recognition with Data Augmentation

Deep Learning–Based Time-Series Classification for Robotic Inspection of Pipe Condition Using Non-Contact Ultrasonic Testing

Abstract This journal paper explores the application of Deep Learning (DL)-based Time-Series Classification (TSC) algorithms in ultrasonic testing for pipeline inspection. The utility of Electromagnetic Acoustic Transducers (EMAT) as a non-contact ultrasonic testing technique for compact robotic platforms is emphasized, prioritizing computational efficiency in defect detection over pinpoint accuracy. To address limited sample availability, the study conducts benchmarking of four methods to enable comparative evaluation of classification times. The core of the DL-based TSC approach involves training DL models using varied proportions (60%, 80%, and 100%) of the available training dataset. This investigation demonstrates the adaptability of DL-enabled anomaly detection with shifting data sizes, showcasing the AI-driven process's robustness in identifying pipeline irregularities. The outcomes underscore the pivotal role of artificial intelligence (AI) in facilitating semi-accurate but swift anomaly detection, thereby streamlining subsequent focused inspections on pipeline areas of concern. By synergistically integrating EMAT technology and DL-driven TSC, this research contributes to enhancing the precision and near real-time inspection capabilities of pipeline assessment. This investigation collectively highlights the potential of DL networks to revolutionize pipeline inspection by rapidly and accurately analyzing ultrasound waveform data.

An Ensemble of Long Short-Term Memory Networks with an Attention Mechanism for Upper Limb Electromyography Signal Classification

Advancing cutting-edge techniques to accurately classify electromyography (EMG) signals are of paramount importance given their extensive implications and uses. While recent studies in the literature present promising findings, a significant potential still exists for substantial enhancement. Motivated by this need, our current paper introduces a novel ensemble neural network approach for time series classification, specifically focusing on the classification of upper limb EMG signals. Our proposed technique integrates long short-term memory networks (LSTM) and attention mechanisms, leveraging their capabilities to achieve accurate classification. We provide a thorough explanation of the architecture and methodology, considering the unique characteristics and challenges posed by EMG signals. Furthermore, we outline the preprocessing steps employed to transform raw EMG signals into a suitable format for classification. To evaluate the effectiveness of our proposed technique, we compare its performance with a baseline LSTM classifier. The obtained numerical results demonstrate the superiority of our method. Remarkably, the method we propose attains an average accuracy of 91.5%, with all motion classifications surpassing the 90% threshold.

Distance-based one-class time-series classification approach using local cluster balance

Deciding the signal length is an important challenge for one-class time-series classification (OCTSC). This paper aims to develop an OCTSC algorithm that does not require model retraining for different signal lengths. For this purpose, a distance-based one-class time-series classification approach using local cluster balance (OCLCB) is proposed. OCLCB extracts feature vectors, namely, local cluster balance (LCB), from the clustering results of sliding windows. K-means clustering is applied to the sliding windows extracted from the training signal. Then, the local prototype (LP) is calculated as the average of the local cluster balance (LCB) in the training data. Unseen scores are computed as the distance metrics between LP and LCBs in the testing data. Since the sliding window size is independent of the entire signal size, OCLCB does not need to retrain the model. This aspect gives the benefit of reducing the parameter tuning costs. The source code is uploaded at https://github.com/ToshiHayashi/OCLCB.

Online Evidential Nearest Neighbour Classification for Internet of Things Time Series

SummaryThe ‘Internet of Things’ (IoT) is a rapidly developing set of technologies that leverages large numbers of networked sensors, to relay data in an online fashion. Typically, knowledge of the sensor environment is incomplete and subject to changes over time. There is a need to employ classification algorithms to understand the data. We first review of existing time series classification (TSC) approaches, with emphasis on the well‐known k‐nearest neighbours (kNN) methods. We extend these to dynamical kNN classifiers, and discuss their shortcomings for handling the inherent uncertainty in IoT data. We next review evidential kNN ( ) classifiers that leverage the well‐known Dempster–Shafer theory to allow principled uncertainty quantification. We develop a dynamic approach for classifying IoT streams via algorithms that use evidential theoretic pattern rejection rules for (i) classifying incoming patterns into a set of oracle classes, (ii) automatically pruning ambiguously labelled patterns such as aberrant streams (due to malfunctioning sensors, say), and (iii) identifying novel classes that may emerge in new subsequences over time. While these methods have wide applicability in many domains, we illustrate the dynamic and approaches for classifying a large, noisy IoT time series dataset from an insurance firm.

CEEMD-MultiRocket: Integrating CEEMD with Improved MultiRocket for Time Series Classification

Time series classification (TSC) is always a very important research topic in many real-world application domains. MultiRocket has been shown to be an efficient approach for TSC, by adding multiple pooling operators and a first-order difference transformation. To classify time series with higher accuracy, this study proposes a hybrid ensemble learning algorithm combining Complementary Ensemble Empirical Mode Decomposition (CEEMD) with improved MultiRocket, namely CEEMD-MultiRocket. Firstly, we utilize the decomposition method CEEMD to decompose raw time series into three sub-series: two Intrinsic Mode Functions (IMFs) and one residue. Then, the selection of these decomposed sub-series is executed on the known training set by comparing the classification accuracy of each IMF with that of raw time series using a given threshold. Finally, we optimize convolution kernels and pooling operators, and apply our improved MultiRocket to the raw time series, the selected decomposed sub-series and the first-order difference of the raw time series to generate the final classification results. Experiments were conducted on 109 datasets from the UCR time series repository to assess the classification performance of our CEEMD-MultiRocket. The extensive experimental results demonstrate that our CEEMD-MultiRocket has the second-best average rank on classification accuracy against a spread of the state-of-the-art (SOTA) TSC models. Specifically, CEEMD-MultiRocket is significantly more accurate than MultiRocket even though it requires a relatively long time, and is competitive with the currently most accurate model, HIVE-COTE 2.0, only with 1.4% of the computing load of the latter.

Using Deep Convolutional Neural Networks to Classify the Discharge Current of a Cold Atmospheric-Pressure Plasma Jet

Atmospheric-pressure plasma discharge can vary with conditions, such as working gas or discharge type, and the resulting discharge current has quite different electrical features. Hence, the provision of real-time and cost-effective monitoring of atmospheric-pressure plasma discharge is possible via current classification with a deep learning model. In addition, the current generated by an atmospheric-pressure plasma jet (APPJ) can be easily obtained, so it is suitable to use a data-driven convolutional neural network (CNN) to extract and analyze the current characteristics. This study presents two CNN classification applications of the current generated by an APPJ. First, we used the time-series classification approach known as InceptionTime to predict the APPJ working gas from helium, argon, and nitrogen. Second, InceptionTime was also used to predict the APPJ discharge type as Townsend discharge or glow discharge. InceptionTime leveraged the use of the inception module to reduce dimensions and avoid overfitting. Moreover, the use of four different sizes of filters along with the maxpooling layer provided sufficient receptive fields to analyze the entire current signal. We achieved 100% accuracy on a testing set of 1125 current signals, which did not include the training data during the working gas prediction; we also achieved 86.3% accuracy on a testing set of 750 current signals, which did not include the training data during the discharge type prediction. This study proves that the CNN can extract features from the discharge current waveforms and successfully predict the discharge type and working gas of an APPJ. The high-speed computing characteristics of deep learning may also facilitate real-time monitoring or diagnosis of APPJ.

Fuzzy Cognitive Map-Driven Comprehensive Time-Series Classification.

This article presents a comprehensive approach for time-series classification. The proposed model employs a fuzzy cognitive map (FCM) as a classification engine. Preprocessed input data feed the employed FCM. Map responses, after a postprocessing procedure, are used in the calculation of the final classification decision. The time-series data are staged using the moving-window technique to capture the time flow in the training procedure. We use a backward error propagation algorithm to compute the required model hyperparameters. Four model hyperparameters require tuning. Two are crucial for the model construction: 1) FCM size (number of concepts) and 2) window size (for the moving-window technique). Other two are important for training the model: 1) the number of epochs and 2) the learning rate (for training). Two distinguishing aspects of the proposed model are worth noting: 1) the separation of the classification engine from pre- and post-processing and 2) the time flow capture for data from concept space. The proposed classifier joins the key advantage of the FCM model, which is the interpretability of the model, with the superior classification performance attributed to the specially designed pre- and postprocessing stages. This article presents the experiments performed, demonstrating that the proposed model performs well against a wide range of state-of-the-art time-series classification algorithms.

End-to-end learned early classification of time series for in-season crop type mapping

Remote sensing satellites capture the cyclic dynamics of our Planet in regular time intervals recorded in satellite time series data. End-to-end trained deep learning models use this time series data to make predictions at a large scale, for instance, to produce up-to-date crop cover maps. Most time series classification approaches focus on the accuracy of predictions. However, the earliness of the prediction is also of great importance since coming to an early decision can make a crucial difference in time-sensitive applications.In this work, we present an End-to-End Learned Early Classification of Time Series (ELECTS) model that estimates a classification score and a probability of whether sufficient data has been observed to come to an early and still accurate decision. ELECTS is modular: any deep time series classification model can adopt the ELECTS conceptual idea by adding a second prediction head that outputs a probability of stopping the classification. The ELECTS loss function then optimizes the overall model on a balanced objective of earliness and accuracy. Our experiments on four crop classification datasets from Europe and Africa show that ELECTS allows reaching state-of-the-art accuracy while reducing the quantity of data massively to be downloaded, stored, and processed. The source code is available at https://github.com/marccoru/elects.

Tracking changes in coastal land cover in the Yellow Sea, East Asia, using Sentinel-1 and Sentinel-2 time-series images and Google Earth Engine

Coastal zones are essential ecosystems due to their provision of invaluable ecosystem services. However, the geomorphologic characteristics of coastal zones are becoming more complex and changeable due to global warming, sea-level rise (SLR), and the intensification of anthropogenic activities. Therefore, accurate and timely knowledge of coastal land cover types (including tidal flats, coastal vegetation, and year-long water cover) is needed for coastal research and sustainable management. To date, land cover products for coastal areas are mainly derived from moderate resolution imaging spectroradiometer images, but few studies have used Sentinel-1 synthetic aperture radar (S1) and Sentinel-2 Multispectral Instrument (S2) images, which can provide more detailed maps. We developed a Rule-based Time Series Classification (RTSC) approach to map coastal land cover types at a 10 m resolution, combining S1/S2 time-series images (2015–2019) and Google Earth Engine (GEE). These products were developed for the coastal zone of the Yellow Sea (YS), East Asia, which is an essential ecosystem protecting a coastal population of 60 million people from storms and SLR effects. Accuracy assessment showed that the annual maps of coastal land cover had high overall accuracy. The coastal land cover types for the YS in 2019 comprised 3593.42 km2 of tidal flats, 28,506.98 km2 of coastal vegetation, and 5436.92 km2 of coastal year-long water. The interannual dynamics of the coastal land cover area in the YS during 2015–2019 were smaller. This study provides a promising method that combines S1/2 time series, a RTSC approach, and GEE to map coastal land cover areas at large scales. The 10 m resolution maps generated in this study are the most current dataset of coastal land cover types for the YS, and they potentially provide a basis for the sustainable management and conservation of this important coastal zone.

Automated Interictal Epileptiform Discharge Detection from Scalp EEG Using Scalable Time-series Classification Approaches.

Deep learning for automated interictal epileptiform discharge (IED) detection has been topical with many published papers in recent years. All existing works viewed EEG signals as time-series and developed specific models for IED classification; however, general time-series classification (TSC) methods were not considered. Moreover, none of these methods were evaluated on any public datasets, making direct comparisons challenging. This paper explored two state-of-the-art convolutional-based TSC algorithms, InceptionTime and Minirocket, on IED detection. We fine-tuned and cross-evaluated them on a public (Temple University Events - TUEV) and two private datasets and provided ready metrics for benchmarking future work. We observed that the optimal parameters correlated with the clinical duration of an IED and achieved the best area under precision-recall curve (AUPRC) of 0.98 and F1 of 0.80 on the private datasets, respectively. The AUPRC and F1 on the TUEV dataset were 0.99 and 0.97, respectively. While algorithms trained on the private sets maintained their performance when tested on the TUEV data, those trained on TUEV could not generalize well to the private data. These results emerge from differences in the class distributions across datasets and indicate a need for public datasets with a better diversity of IED waveforms, background activities and artifacts to facilitate standardization and benchmarking of algorithms.

LSTM-Based Intrusion Detection System for VANETs: A Time Series Classification Approach to False Message Detection

In vehicular ad hoc networks (VANETs), vehicles broadcast emergency messages and beacon messages, which enable drivers to perceive traffic conditions beyond their visual range thus improve driving safety. However, internal attackers can launch a false message attack for selfish purposes by reporting a non-existent traffic incident in emergency messages. Moreover, some collusion attackers may spread bogus beacon messages cooperatively to make the bogus traffic incident more deceptive. To improve the accuracy of false emergency message detection, we propose a novel intrusion detection system (IDS) based on time series classification and deep learning. Considering that traffic parameters are highly correlated with time, we collect time series of traffic parameters closely related to traffic incidents from messages of vehicles near reported traffic incidents as time series feature vectors. To recognize the pattern of traffic parameters changing over time more accurately, a traffic incident classifier based on long short-term memory (LSTM) is designed and trained using time series feature vectors from both normal and collusion attack scenarios. Based on the classification result, the authenticity of the emergency message can be determined. Finally, we evaluate the performance of the proposed LSTM-based IDS through extensive simulation. Simulation results validate that our IDS is more accurate in false message detection compared with some well-known machine learning-based schemes.

Deep-Learning-Based Multivariate Time-Series Classification for Indoor/Outdoor Detection

Recently, the topic of indoor outdoor detection (IOD) has seen its popularity increase, as IOD models can be leveraged to augment the performance of numerous Internet of Things and other applications. IOD aims at distinguishing in an efficient manner whether a user resides in an indoor or an outdoor environment, by inspecting the cellular phone sensor recordings. Legacy IOD models attempt to determine a user’s environment by comparing the sensor measurements to some threshold values. However, as we also observe in our experiments, such models exhibit limited scalability, and their accuracy can be poor. Machine learning (ML)-based IOD models aim at removing this limitation, by utilizing a large volume of measurements to train ML algorithms to classify a user’s environment. Yet, in most of the existing research, the temporal dimension of the problem is disregarded. In this article, we propose treating IOD as a multivariate time-series classification (TSC) problem, and we explore the performance of various deep learning (DL) models. We demonstrate that a multivariate TSC approach can be used to monitor a user’s environment, and predict changes in its state, with greater accuracy compared to conventional approaches that ignore the feature variation over time. Additionally, we introduce a new DL model for multivariate TSC, exploiting the concept of self-attention and atrous spatial pyramid pooling. The proposed DL multivariate TSC framework exploits only low power consumption sensors to infer a user’s environment, and it outperforms state-of-the-art models, yielding a higher accuracy combined with a smaller computational cost.

Similarity-based Multi-source Transfer Learning Approach for Time Series Classification

This study aims to develop an effective method of classification concerning time series signals for machine state prediction to advance predictive maintenance (PdM). Conventional machine learning (ML) algorithms are widely adopted in PdM, however, most existing methods assume that the training (source) and testing (target) data follow the same distribution, and that labeled data are available in both source and target domains. For real-world PdM applications, the heterogeneity in machine original equipment manufacturers (OEMs), operating conditions, facility environment, and maintenance records collectively lead to heterogeneous distribution for data collected from different machines. This will significantly limit the performance of conventional ML algorithms in PdM. Moreover, labeling data is generally costly and time-consuming. Finally, industrial processes incorporate complex conditions, and unpredictable breakdown modes lead to extreme complexities for PdM. In this study, similarity-based multi-source transfer learning (SiMuS-TL) approach is proposed for real-time classification of time series signals. A new domain, called "mixed domain," is established to model the hidden similarities among the multiple sources and the target. The proposed SiMuS-TL model mainly includes three key steps: 1) learning group-based feature patterns, 2) developing group-based pre-trained models, and 3) weight transferring. The proposed SiMuS-TL model is validated by observing the state of the rotating machinery using a dataset collected on the Skill boss manufacturing system, publicly available standard bearing datasets, Case Western Reserve University (CWRU), and Paderborn University (PU) bearing datasets. The results of the performance comparison demonstrate that the proposed SiMuS-TL method outperformed conventional Support Vector Machine (SVM), Artificial Neural Network (ANN), and Transfer learning with neural networks (TLNN) without similarity-based transfer learning methods.

OCSTN: One-class time-series classification approach using a signal transformation network into a goal signal

One-class classification (OCC) is a classification task where the training data have only one class. The goal is to classify input data into one seen class or other unseen classes. This paper proposes an OCC approach using a signal transformation network (OCSTN), which aims to process univariate time-series data. The main contribution is developing a signal transformation network (STN) that aims to transform input signals into one signal, namely the goal signal. Moreover, the model error of the STN is a distance metric between the goal signal and the model output. The STN model learns from one-class signals. Therefore, model error for one class is small relative to other classes. Accordingly, OCSTN could discriminate between seen and unseen classes using the model errors. The proposed OCSTN is evaluated using two ballistocardiography (BCG) datasets. The OCSTN achieves fair results in both AUC scores and processing speed. OCSTN has a weak point in training diverse signals. In addition, the entropy and smoothness of the goal signal are highly related to the AUC score.

Learning the micro-environment from rich trajectories in the context of mobile crowd sensing

With the rapid advancements of sensor technologies and mobile computing, Mobile Crowd Sensing (MCS) has emerged as a new paradigm to collect massive-scale rich trajectory data. Nomadic sensors empower people and objects with the capability of reporting and sharing observations on their state, their behavior and/or their surrounding environments. Processing and mining multi-source sensor data in MCS raise several challenges due to their multi-dimensional nature where the measured parameters (i.e., dimensions) may differ in terms of quality, variability, and time scale. We consider the context of air quality MCS and focus on the task of mining the micro-environment from the MCS data. Relating the measures to their micro-environment is crucial to interpret them and analyse the participant’s exposure properly. In this paper, we focus on the problem of investigating the feasibility of recognizing the human’s micro-environment in an environmental MCS scenario. We propose a novel approach for learning and predicting the micro-environment of users from their trajectories enriched with environmental data represented as multidimensional time series plus GPS tracks. We put forward a multi-view learning approach that we adapt to our context, and implement it along with other time series classification approaches. We extend the proposed approach to a hybrid method that employs trajectory segmentation to bring the best of both methods. We optimise the proposed approaches either by analysing the exact geolocation (which is privacy invasive), or simply applying some a priori rules (which is privacy friendly). The experimental results, applied to real MCS data, not only confirm the power of MCS and air quality (AQ) data in characterizing the micro-environment, but also show a moderate impact of the integration of mobility data in this recognition. Furthermore, and during the training phase, multi-view learning shows similar performance as the reference deep learning algorithm, without requiring specific hardware. However, during the application of models on new data, the deep learning algorithm fails to outperform our proposed models.

KDCTime: Knowledge distillation with calibration on InceptionTime for time-series classification

Time-series classification approaches based on deep neural networks easily overfit UCR datasets, which is caused by the few-shot problem of those datasets. Therefore, to alleviate the overfitting phenomenon to further improve accuracy, we first propose label smoothing for InceptionTime (LSTime), which adopts the soft label information compared to only hard labels. Next, instead of manually adjusting soft labels by LSTime, knowledge distillation for InceptionTime (KDTime) is proposed to automatically generate soft labels by the teacher model while compressing the inference model. Finally, to rectify the incorrectly predicted soft labels from the teacher model, knowledge distillation with calibration for InceptionTime (KDCTime) is proposed, which contains two optional calibrating strategies, i.e., KDC by translating (KDCT) and KDC by reordering (KDCR). The experimental results show that the KDCTime accuracy is promising, while its inference time is orders of magnitude faster than state-of-the-art approaches.

FedTSC

We demonstrate FedTSC, a novel federated learning (FL) system for interpretable time series classification (TSC). FedTSC is an FL-based TSC solution that makes a great balance among security, interpretability, accuracy, and efficiency. We achieve this by first extending the concept of FL to consider both stronger security and model interpretability. Then, we propose three novel TSC methods based on explainable features to deal with the challengeable FL problem. To build the model in the FL setting, we propose several security protocols that are well optimized by maximally reducing the bottlenecked communication complexity. We build the FedTSC system based on such a solution, and provide the user Sklearn-like Python APIs for practical utility. We show that the system is easy to use, and the novel TSC approach is superior.