Time Series Classification Problem Research Articles

Discriminate Supervised Weighted Scheme for the Classification of Time Series Signals

Innovations and applicability of time series data mining techniques have significantly increased the researchers' interest in the problem of time series classification. Several algorithms have been proposed for this purpose categorized under shapelet, interval, motif, and whole series-based techniques. Among this, the bag-of-words technique, an extensive application of the text mining approach, performs well due to its simplicity and effectiveness. To extend the efficiency of the bag-of-words technique, this paper proposes a discriminate supervised weighted scheme to identify the characteristic and representative pattern of a class for efficient classification. This paper uses a modified weighted matrix that discriminates the representative and non-representative pattern which enables the interpretability in classification. Experimentation has been carried out to compare the performance of the proposed technique with state-of-the-art techniques in terms of accuracy and statistical significance.

Data Anomaly Detection of Bridge Structures Using Convolutional Neural Network Based on Structural Vibration Signals

Structural monitoring provides valuable information on the state of structural health, which is helpful for structural damage detection and structural state assessment. However, when the sensors are exposed to harsh environmental conditions, various anomalies caused by sensor failure or damage lead to abnormalities of the monitoring data. It is inefficient to remove abnormal data by manual elimination because of the massive number of data obtained by monitoring systems. In this paper, a data anomaly detection method based on structural vibration signals and a convolutional neural network (CNN) is proposed, which can automatically identify and eliminate abnormal data. First, the anomaly detection problem is modeled as a time series classification problem. Data preprocessing and data augmentation, including data expansion and down-sampling to construct new samples, are employed to process the original time series. For a small number of samples in the data set, randomly increase outliers, symmetrical flipping, and noise addition methods are used for data expansion, and samples with the same label are added without increasing the original samples. The down-sampling method of symmetrically extracting the maximum value and the minimum value at the same time can effectively reduce the dimensionality of the input sample, while retaining the characteristics of the data to the greatest extent. Using hyperparameter tuning of the classification weights, CNN is more effective in dealing with unbalanced training sets. Finally, the effectiveness of the proposed method is proved by the anomaly detection of acceleration data on a long-span bridge. For the anomaly detection problem modeled as a time series classification problem, the proposed method can effectively identify various abnormal patterns.

A novel wide & deep transfer learning stacked GRU framework for network intrusion detection

With the increasing frequency, severity and complexity of recent cyber attacks around the world, network intrusion detection has become mandatory and highly sophisticated task. Achieving high performance in network intrusion detection by applying benchmark machine learning classifiers (including deep learning techniques) has become a major challenge in recent times. One of the biggest challenges is improving the memorization capacity and generalization ability of NIDS (Network Intrusion Detection Systems). In this paper, we propose a highly scalable novel wide & deep transfer learning (TL) based stacked GRU (Gated Recurrent Unit) model to deal with multi-dimensional point data and multi-variate time series regression and classification problems in network intrusion detection. The proposed model has the memorization capacity of linear regression model and the generalization ability of deep GRU model. The deep component consists of a transfer learning framework that pretrains a source model and then fine-tunes the whole source model on the same dataset multiple times until it gives peak performance. This method gives a multi-class evaluation accuracy of 99.92% on KDDCup 99(10%) dataset and 94.22% on UNSW-NB15 dataset respectively. Extensive experimentations and evaluations have been carried out by comparing it with other machine learning (including deep learning) network intrusion detection techniques. The proposed method outperforms most of the existing intrusion detection approaches.

Auto-adaptive multilayer perceptron for univariate time series classification

Time Series Classification (TSC) is an intricate problem that has encountered applications in various science fields. Accordingly, many researchers have presented interesting proposals to tackle the TSC problem. Nevertheless, most methods are hand-crafted to classify specific Time Series (TS) and are computationally expensive even for small data sets. In this paper, we propose a new approach to the Multilayer Perceptron (MLP) for TSC. The main novelty is that the hyperparameters related to batch size and the number of neurons in the hidden layers are auto-adapted according to the TS nature. We carried out an empirical study on 61 benchmark data sets from the University of California, Riverside (UCR). The experimental evaluation revealed that our proposal is competitive when we compare the accuracy versus 14 state-of-the-art methods. A non-parametric statistical test verifies that the proposed MLP ranked in fourth place and can be executed on standard computer equipment, making it simple, accessible, and competitive.

Echo state network with a global reversible autoencoder for time series classification

• In order to reestablish the random initialization input weights of ESN, a single layer ESN with a global reversible AE (GRAE) algorithm is proposed to ensure that the outputs of GRAE are remarkably correlated with the input data. • The reservoir layer with a reversible function is calculated by pulling the ESN output back and injecting it into the reservoir layer. Thus, the feature learning is enriched by additional information, which results in a better performance. • The GRAE which is based on ridge regression to solve a least squares problem is entirely different from existing back-propagation (BP) based AEs. So, the training speed of GRAE is many times faster than that of the BP based AEs. • In existing ESN autoencoder methods, all weights in the reservoir layer are generated randomly. The simple cycle connection is utilized in the reservoir layer of ESN to make sure the reservoir weights can be generated deterministically. An echo state network (z) can provide an efficient dynamic solution for predicting time series problems. However, in most cases, ESN models are applied for predictions rather than classifications. The applications of ESN in time series classification (TSC) problems have yet to be fully studied. Moreover, the conventional randomly generated ESN is unlikely to be optimal because of the randomly generated input and reservoir weights, which are not always guaranteed to be optimal. Randomly generating all layer weights is improper, because a purely random layer might destroy the useful features. To overcome this disadvantage, this study provides a new input weight establishment framework of ESN based on autoencoder (AE) theory for TSC tasks. A global reversible AE (GRAE) algorithm is proposed to reestablish the random initialization input weights of the ESN. In existing ESN-AEs, the output weights obtained in the encoding process are directly reused as the initial input weights. By contrast, in GRAE, the reservoir layer with a reversible activation function is calculated by pulling the decoding layer output back and injecting it into the reservoir layer. Thus, feature learning is enriched by additional information, which results in improved performance. The current weights of the encoding layer are iteratively replaced by the decoding layer to ensure that the outputs of the GRAE are remarkably correlated with the input data. Visualization analyses and experiments of the input weights on a massive set of UCR time series datasets indicate that the proposed GRAE method can considerably improve the original two-layer ESN-based classifiers and the proposed GRAE-ESN classifier yields better performance compared with traditional state-of-the-art TSC classifiers. Furthermore, the proposed method can provide comparable performance and considerably faster training speed compared with three deep learning classifiers.

Feature extraction by grammatical evolution for one-class time series classification

When dealing with a new time series classification problem, modellers do not know in advance which features could enable the best classification performance. We propose an evolutionary algorithm based on grammatical evolution to attain a data-driven feature-based representation of time series with minimal human intervention. The proposed algorithm can select both the features to extract and the sub-sequences from which to extract them. These choices not only impact classification performance but also allow understanding of the problem at hand. The algorithm is tested on 30 problems outperforming several benchmarks. Finally, in a case study related to subject authentication, we show how features learned for a given subject are able to generalise to subjects unseen during the extraction phase.

“A new feature-based time series classification method by using scale-space extrema”

Time series data mining has received significant attention over the past decade, and many approaches have focused on classification tasks where the goal is to define the label of a test time series, given labeled training data. Time series classification approaches can be broadly grouped into two categories as instance-based and feature-based methods. Instance-based approaches utilize similarity information in a nearest-neighbor setting to classify time series data. Although approaches from this category provide accurate results, their performance degrades with long and noisy time series. On the other hand, feature-based approaches extract features to deal with the limitations of instance-based approaches; however, these approaches work with predefined features and may not be successful in certain classification problems. This study proposes a time series classification approach that benefits from both scale-space theory and bag-of-features technique. The method starts with finding the scale-space extrema points (i.e. key points) of each time series according to the SiZer (SIgnificant ZERo crossings of the derivatives) method, and then proceeds to create the local features set around these points. After extraction of the local features from each key point, a bag-of-features representation for each time series is constructed as the summary of the key point characteristics. We evaluate the success of the proposed representation on time series classification problems from various domains. Our experimental results show that our proposal provides competitive results compared to widely used approaches in the literature.

Mo-BoNet: A TIME SERIES CLASSIFICATION MODEL BASED ON COMPUTER VISION

Time series are widely distributed in many fields. Classical statistical methods are difficult to model the deep meaning of time series, and the deep learning methods based on recurrent neural network has great limitations when it is applied to indefinite long time series. In order to solve the above problems, a time series classification model based on computer vision is proposed, which transforms the time series classification problem into image classification problem. Firstly, three kinds of images with different linewidth corresponding to the time series are used as input to reduce the information loss in the conversion process. Secondly, the transfer learning model based on MobileNetV3-Large is used to encode the image data, and XGBoost is used for classification. The experimental results show that the classification effect of this model is better than that of the classical image classification model, and its XGBoost is also better than other ensemble methods, which proves the feasibility of computer vision method in time series classification task.

Attention-Based Deep Gated Fully Convolutional End-to-End Architectures for Time Series Classification

Time series classification (TSC) is one of the significant problems in the data mining community due to the wide class of domains involving the time series data. The TSC problem is being studied individually for univariate and multivariate using different datasets and methods. Subsequently, deep learning methods are more robust than other techniques and revolutionized many areas, including TSC. Therefore, in this study, we exploit the performance of attention mechanism, deep Gated Recurrent Unit (dGRU), Squeeze-and-Excitation (SE) block, and Fully Convolutional Network (FCN) in two end-to-end hybrid deep learning architectures, Att-dGRU-FCN and Att-dGRU-SE-FCN. The performance of the proposed models is evaluated in terms of classification testing error and f1-score. Extensive experiments and ablation study is carried out on multiple univariate and multivariate datasets from different domains to acquire the best performance of the proposed models. The proposed models show effective performance over other published methods, also do not require heavy data pre-processing, and small enough to be deployed on real-time systems.

SRPM\u2013CNN: a combined model based on slide relative position matrix and CNN for time series classification

Research on the time series classification is gaining an increased attention in the machine learning and data mining areas due to the existence of the time series data almost everywhere, especially in our daily work and life. Recent studies have shown that the convolutional neural networks (CNN) can extract good features from the images and texts, but it often encounters the problem of low accuracy, when it is directly employed to solve the problem of time series classification. In this pursuit, the present study envisaged a novel combined model based on the slide relative position matrix and CNN for time series. The proposed model first adopted the slide relative position for converting the time series data into 2D images during preprocessing, and then employed CNN to classify these images. This made the best of the temporal sequence characteristic of time series data, thereby utilizing the advantages of CNN in image recognition. Finally, 14 UCR time series datasets were chosen to evaluate the performance of the proposed model, whose results indicate that the accuracy of the proposed model was higher than others.

Multi-scale Attention Convolutional Neural Network for time series classification

With the rapid increase of data availability, time series classification (TSC) has arisen in a wide range of fields and drawn great attention of researchers. Recently, hundreds of TSC approaches have been developed, which can be classified into two categories: traditional and deep learning based TSC methods. However, it remains challenging to improve accuracy and model generalization ability. Therefore, we investigate a novel end-to-end model based on deep learning named as Multi-scale Attention Convolutional Neural Network (MACNN) to solve the TSC problem. We first apply the multi-scale convolution to capture different scales of information along the time axis by generating different scales of feature maps. Then an attention mechanism is proposed to enhance useful feature maps and suppress less useful ones by learning the importance of each feature map automatically. MACNN addresses the limitation of single-scale convolution and equal weight feature maps. We conduct a comprehensive evaluation of 85 UCR standard datasets and the experimental results show that our proposed approach achieves the best performance and outperforms the other traditional and deep learning based methods by a large margin.

Depth Evaluation for Metal Surface Defects by Eddy Current Testing Using Deep Residual Convolutional Neural Networks

Eddy current testing (ECT) is an effective technique for evaluating depth of metal surface defects. However, in practice, evaluation primarily relies on the experience of an operator and is often carried out by manual inspection. In this article, we address the challenges of automatic depth evaluation of metal surface defects by virtual of state-of-the-art deep learning (DL) techniques. The main contributions are threefold. First, a highly integrated portable ECT device is developed, taking the advantage of an advanced field-programmable gate array (Zynq-7020 system on chip) and provides fast data acquisition and in-phase/quadrature demodulation. Second, a dataset, termed metal defects of different depths by ECT (MDDECT), is constructed using the ECT device by human operators and made openly available. It contains 48000 scans from 18 defects of different depths and liftoffs. Third, the depth evaluation problem is formulated as a time series classification problem, and various state-of-the-art 1-D residual convolutional neural networks are trained and evaluated on the MDDECT dataset. A 38-layer 1-D ResNeXt achieves an accuracy of 93.58% in discriminating the surface defects in a stainless steel sheet with depths from 0.3 to 2.0 mm in the resolution of 0.1 mm. In addition, the results show that the trained ResNeXt1D-38 model is immune to liftoff signals.

Performance Boosting of Scale and Rotation Invariant Human Activity Recognition (HAR) with LSTM Networks Using Low Dimensional 3D Posture Data in Egocentric Coordinates

Human activity recognition (HAR) has been an active area in computer vision with a broad range of applications, such as education, security surveillance, and healthcare. HAR is a general time series classification problem. LSTMs are widely used for time series classification tasks. However, they work well with high-dimensional feature vectors, which reduce the processing speed of LSTM in real-time applications. Therefore, dimension reduction is required to create low-dimensional feature space. As it is experimented in previous study, LSTM with dimension reduction yielded the worst performance among other classifiers, which are not deep learning methods. Therefore, in this paper, a novel scale and rotation invariant human activity recognition system, which can also work in low dimensional feature space is presented. For this purpose, Kinect depth sensor is employed to obtain skeleton joints. Since angles are used, proposed system is already scale invariant. In order to provide rotation invariance, body relative direction in egocentric coordinates is calculated. The 3D vector between right hip and left hip is used to get the horizontal axis and its cross product with the vertical axis of global coordinate system assumed to be the depth axis of the proposed local coordinate system. Instead of using 3D joint angles, 8 number of limbs and their corresponding 3D angles with X, Y, and Z axes of the proposed coordinate system are compressed with several dimension reduction methods such as averaging filter, Haar wavelet transform (HWT), and discrete cosine transform (DCT) and employed as the feature vector. Finally, extracted features are trained and tested with LSTM (long short-term memory) network, which is an artificial recurrent neural network (RNN) architecture. Experimental and benchmarking results indicate that proposed framework boosts the performance of LSTM by approximately 30% accuracy in low-dimensional feature space.

Deep Learning for Heterogeneous Human Activity Recognition in Complex IoT Applications

With continued improvements in wireless sensing technology, the notion of the Internet of Things (IoT) has been widely adopted and has become pervasive owing to its broad applications in scenarios such as ambient assisted living, smart healthcare, and smart homes. In that regard, human activity recognition (HAR) is a vital element of intelligent systems to undertake persistent surveillance of human behavior. Due to the omnipresent impact of smartphones in each person’s life, smartphone inertial sensors are used as a case study for this research. Most of the conventional approaches regard HAR as a time-series classification problem; yet, the accuracy of recognition degrades for heterogeneous sensors. In this article, we investigate encoding sensory heterogeneous HAR (HHAR) data into three-channel image representation (i.e., RGB), hence treat the HHAR task as an image classification problem. Since present convolutional network models are computationally heavy when deployed in the IoT environment, we propose a lightweight model image encoded HHAR, called multiscale image-encoded HHAR (MS-IE-HHAR). The model employs a hierarchical multiscale extraction (HME) module followed by an improved spatialwise and channelwise attention (ISCA) module to form the main architecture of the model. The HME module is formed by a group of residually connected shuffle group convolutions (SG-Conv) to extract and learn image representations from different receptive fields while reducing the number of network parameters. The ISCA module combines a lightweight spatialwise attention (SwA) block and an improved channelwise attention (CwA) module to enable the network to pay instructive attention to spatial correlations as well as channel interdependency information. Finally, two widely available HHAR public data sets (i.e., HHAR UCI, and MHEALTH) were used to evaluate the performance of the proposed models with accuracy over 98% and 99%, respectively, demonstrating the model superiority for modeling HAR from heterogeneous data sources.

A Photonic Recurrent Neuron for Time-Series Classification

Neuromorphic photonics has turned into a key research area for enabling neuromorphic computing at much higher data-rates compared to their electronic counterparts, improving significantly the (multiply-and-accumulate) MAC/sec. At the same time, time-series classification problems comprise a large class of artificial intelligence (AI) applications where speed and latency can have a decisive role in their hardware deployment roadmap, highlighting the need for ultra-fast hardware implementations of simplified recurrent neural networks (RNN) that can be extended in more advanced long-short-term-memory (LSTM) and gated recurrent unit (GRU) machines. Herein, we experimentally demonstrate a novel photonic recurrent neuron (PRN) to classify successfully a time-series vector with 100-psec optical pulses and up to 10 Gb/s data speeds, reporting on the fastest all-optical real-time classifier. Experimental classification of 3-bit optical binary data streams is presented, revealing an average accuracy of >91% and confirming the potential of PRNs to boost speed and latency performance in time-series AI applications.

A Deep Learning Approach to Detect and Isolate Thruster Failures for Dynamically Positioned Vessels Using Motion Data

Vessels today are being fully monitored, thanks to the advance of sensor technology. The availability of data brings ship intelligence into great attention. As part of ship intelligence, the desire of using advanced data-driven methods to optimize operation also increases. Considering ship motion data reflects the dynamic positioning performance of the vessels and thruster failure might cause drift-offs, it is possible to detect and isolate potential thruster failure using motion data. In this article, thruster failure detection and isolation are considered as a time-series classification problem. A convolutional neural network (CNN) is introduced to learn the mapping from the logged motion sequence to the status of the thruster. CNN is expected to generate task-specific features from the original time series sensors data and then perform the classification. The data set is collected from a professional simulator in the Offshore Simulation Center AS. Experiments show that the proposed method can detect and isolate failed thrusters with up to 95% accuracy. The proposed model is further extended to deal with thruster failure in a real-time manner.

A Support Vector Machine Classification-Based Signal Detection Method in Ultrahigh-Frequency Radio Frequency Identification Systems

Noise and distortion in real signals have posed challenges in signal detection in ultrahigh-frequency (UHF) radio frequency identification (RFID) systems, with the exception of large frequency deviation. In this article, signal detection in a UHF RFID system was treated as a time-series classification problem. A support vector machine-based detector was proposed for real-time signal detection to optimize the system performance. The improved system was able to achieve frame synchronization detection and time-series data detection with a frequency deviation of ± 22%. With the analysis of the frame format, the frame synchronization was based on the sequence detection with the most likely selection, and the data detection was performed using a symbol-by-symbol method with two-step adjustments in order to reduce time and hardware costs. The algorithm was mainly trained offline using the experimental data, and then efficiently implemented in a UHF RFID system based on a software-defined radio platform. The system was validated on real-time signals of commercial tags. The experimental results showed that the learned detector had displayed improvements in detection accuracy and stability when compared to the existing method based on a correlation algorithm.

Interval Feature Transformation for Time Series Classification Using Perceptually Important Points

A novel feature reconstruction method, referred to as interval feature transformation (IFT), is proposed for time series classification. The IFT uses perceptually important points to segment the series dynamically into subsequences of unequal length, and then extract interval features from each time series subsequence as a feature vector. The IFT distinguishes the best top-k discriminative feature vectors from a data set by information gain. Utilizing these discriminative feature vectors, transformation is applied to generate new k-dimensional data which are lower-dimensional representations of the original data. In order to verify the effectiveness of this method, we use the transformed data in conjunction with some traditional classifiers to solve time series classification problems and make comparative experiments to several state-of-the-art algorithms. Experiment results verify the effectiveness, noise robustness and interpretability of the IFT.

A systematic approach to multi-task learning from time-series data

There have been significant advances in machine learning due to the profusion in data collection and computing resources. However, the need for large annotated datasets to train machine learning models remains a problematic constraint. To address the limitation of annotated data for personalized prediction, we propose a framework to enrich annotated time-series (TS) sensing data by way of transfer learning with new multi-task learning (MTL) models. Compared to previous MTL approaches for TS, this work introduces three contributions. First, we propose a systematic method to examine the efficiency of MTL approaches by exploring options for three key characteristics of MTL with TS: the choice of features that efficiently capture temporally dynamic information, the similarity measure that effectively models the commonality and uniqueness across tasks being learned, and the choice of regularization for achieving the best tradeoff between a model’s generalizability and accuracy. Second, we present an MTL deep learning model that is shown to achieve state-of-the-art performance for personalized human activity recognition from time-series. Experimental results on three benchmark activity recognition datasets and one activity recognition in-the-wild dataset show that the proposed framework provides performance gains over prior work while presenting a unified approach for designing MTL solutions for personalized time-series classification problems.

Time-series classification for industrial applications: a brake pad wear prediction use case

Brake system is an important part for control of a vehicle. Hence condition monitoring of brake pads is essential for ensuring passenger’s safety. Many existing methods for brake pads wear assessment rely on specific sensors installed in the brake system, which could be expensive. Instead we use data from existing vehicle’s sensors and electronic control unit that are readily available in modern vehicles. We reduced the prediction problem to time-series classification problem and developed and tested several classification pipelines based on machine learning. We demonstrated that it is possible to predict a brake pad wear with an accuracy sufficient for real-life usage.