Time Series Mining Techniques Research Articles

Accelerating time series similarity search under Move-Split-Merge distance via dissimilarity space embedding

The ubiquity sensory devices across scientific and industrial settings have made time-series data mining an important research topic in the IoT domain. Among all time-series mining techniques, the similarity search is considered the most essential and noteworthy because it is frequently employed as a fundamental subroutine for other procedures. The Dynamic Time Warping (DTW) was once considered the best similarity measure in time-series data mining community; however, two recent studies have demonstrated that the Move-Split-Merge (MSM) performs better than DTW in both supervised and unsupervised settings. Computing MSM between two time-series incurs costly computation because MSM has quadratic time complexity. Regrettably, there is little research on accelerating time-series similarity search under MSM. This paper provides an efficient method named DSEMSM for speeding up this procedure. DSEMSM utilizes dissimilarity space embedding to transform time-series into low-dimensional vectors. The filter-and-verification framework is subsequently employed to reduce the large number of actual MSM computations and to identify the final search results from the candidate set. The furthest prototype selection strategy and lower bound based technique are applied in DSEMSM to improve its retrieval accuracy and efficiency, respectively. We validate the effectiveness of the proposed method on a large number of datasets. The experimental results indicate that, in comparison to its relevant competitors, DSEMSM achieves a more stable speedup rate and higher retrieval accuracy.

Time Series Shapelet-Based Movement Intention Detection Toward Asynchronous BCI for Stroke Rehabilitation

Brain computer interface (BCI) systems for neurorehabilitation have received increasing attention over the past decade. These systems provide an alternative approach to restore lost motor functions in stroke patients by inducing brain plasticity. To utilize BCI systems for stroke rehabilitation, user movement intentions from electroencephalogram are used as triggers for rehabilitation tools such as electrical stimulators. A reliable movement intention detection plays a vital role in an effective rehabilitation to establish a relation between movement intentions and corresponding feedback from a rehabilitation tool. We therefore propose a novel movement intention detection algorithm based on time series shapelets. We collected a dataset of subjects performing a self-pace ankle dorsiflexion and tested the algorithm in both classification and pseudo-online detection. The classification results demonstrate that our proposed algorithm significantly outperforms six other algorithms and achieves the highest classification performance with F1-score of 0.82. For pseudo-online detection, our algorithm gains very high performance in all performance metrics with 69% True Positive Rate (TPR), 8 False Positives per minute (FPs/min), and 44 ms Detection Latency (DL). Our proposed algorithm does not only provide competitive performance to state-of-the-art algorithms in terms of TPR, but also maintain low FPs/min and DL. In addition, the DL of our proposed algorithm is low enough to induce effective brain plasticity for neurorehabilitation. These promising results enlighten the development of asynchronous BCI systems based on time series mining techniques to enhance stroke rehabilitation results.

Handling Irregularly Sampled IoT Time Series to Inform Infrastructure Asset Management

Infrastructure systems in today's increasingly interconnected world employ the capabilities of the Internet of Things (IoT) technologies for their monitoring, operational control, and asset management. IoT devices can be defined as sensors (of different types) collecting, processing, and sharing time series of data. The analysis of such data often face challenges as a consequence of the high frequency of data collection and the increasing number of sensors placed on infrastructure. Power related issues, timestamp misalignment, and heterogeneous sampling designs are among the most common issues that the IoT data collection may suffer alongside the inherent complexities of large scale databases. This paper provides an overview of time series mining techniques adapted to tackle such issues in IoT data. The aim is to have a pattern recognition tool-set for developing anomaly detection algorithms. Particularly, the paper investigates how to efficiently handle large-scale time series coming from multiple sensors in a stream and following an unevenly spaced - irregular - sampling. The analysis is demonstrated through a case study of time series data mining of sensors installed for supporting the predictive maintenance of quay-cranes at the Port of Felixstowe, the largest container port in Britain.

Об одном методе восстановления пропущенных значений потокового временного ряда в режиме реального времени

The problem of the imputation of missing values in a streaming time series arises in a wide range of Industry 4.0 and Internet of Things applications. In the article, we propose a novel imputation method based on time series mining techniques and artificial neural networks. The method involves three steps of imputation: data preprocessing, recognition, and reconstruction. Preprocessing is a one-time preparation of training data samples. Recognition and reconstruction are implemented through two neural networks trained on the samples above. Preprocessing supposes the discovery of a set of typical subsequences (snippets) in a pre-stored fragment of the streaming time series without misses. Recognition is implemented through a Convolutional Neural Network, and its input is a vector of the elements preceding the current (missing) value. The Recognizer outputs the snippet that the input subsequence is most similar to. Reconstruction is implemented through a Recurrent Neural Network, and its input is a concatenation of the Recognizer's output and the vector of the elements preceding the missing value. The Reconstructor outputs the value to be imputed. The experimental results show high accuracy and the advantage of the proposed method over analogs.

GRAIL

The analysis of time series is becoming increasingly prevalent across scientific disciplines and industrial applications. The effectiveness and the scalability of time-series mining techniques critically depend on design choices for three components responsible for (i) representing; (ii) comparing; and (iii) indexing time series. Unfortunately, these components have to date been investigated and developed independently, often resulting in mutually incompatible methods. The lack of a unified approach has hindered progress towards fast and accurate analytics over massive time-series collections. To address this major drawback, we present GRAIL, a generic framework to learn compact time-series representations that preserve the properties of a user-specified comparison function. Given the comparison function, GRAIL (i) extracts landmark time series using clustering; (ii) optimizes necessary parameters; and (iii) exploits approximations for kernel methods to construct representations in linear time and space by expressing each time series as a combination of the landmark time series. We extensively evaluate GRAIL for querying, classification, clustering, sampling, and visualization of time series. For these tasks, methods leveraging GRAIL's representations are significantly faster and at least as accurate as state-of-the-art methods operating over the raw time series. GRAIL shows promise as a new primitive for highly accurate, yet scalable, time-series analysis.

Measuring the Impact of Financial News and Social Media on Stock Market Modeling Using Time Series Mining Techniques

In this work, we study the task of predicting the closing price of the following day of a stock, based on technical analysis, news articles and public opinions. The intuition of this study lies in the fact that technical analysis contains information about the event, but not the cause of the change, while data like news articles and public opinions may be interpreted as a cause. The paper uses time series analysis techniques such as Symbolic Aggregate Approximation (SAX) and Dynamic Time Warping (DTW) to study the existence of a relation between price data and textual information, either from news or social media. Pattern matching techniques from time series data are also incorporated, in order to experimentally validate potential correlations of price and textual information within given time periods. The ultimate goal is to create a forecasting model that exploits the previously discovered patterns in order to augment the forecasting accuracy. Results obtained from the experimental phase are promising. The performance of the classifier shows clear signs of improvement and robustness within the time periods where patterns between stock price and the textual information have been identified, compared to the periods where patterns did not exist.

Mining lake time series using symbolic representation

Sensor networks deployed in lakes and reservoirs, when combined with simulation models and expert knowledge from the global community, are creating deeper understanding of the ecological dynamics of lakes. However, the amount of data and the complex patterns in the data demand substantial compute resources and efficient data mining algorithms, both of which are beyond the realm of traditional limnological research. This paper uniquely adapts methods from computer science for application to data intensive ecological questions, in order to provide ecologists with approachable methodology to facilitate knowledge discovery in lake ecology. We apply a state-of-the-art time series mining technique based on symbolic representation (SAX) to high-frequency time series of phycocyanin (PHYCO) and chlorophyll (CHLORO) fluorescence, both of which are indicators of algal biomass in lakes, as well as model predictions of algal biomass (MODEL). We use data mining techniques to demonstrate that MODEL predicts PHYCO better than it predicts CHLORO. All time series have high redundancy, resulting in a relatively small subset of unique patterns. However, MODEL is much less complex than either PHYCO or CHLORO and fails to reproduce high biomass periods indicative of algal blooms. We develop a set of tools in R to enable motif discovery and anomaly detection within a single lake time series, and relationship study among multiple lake time series through distance metrics, clustering and classification. Furthermore, to improve computation times, we provision web services to launch R tools remotely on high performance computing (HPC) resources. Comprehensive experimental results on observational and simulated lake data demonstrate the effectiveness of our approach.

Pattern recognition in multivariate time series – A case study applied to fault detection in a gas turbine

Advances in information technology, together with the evolution of systems in control, automation and instrumentation have enabled the recovery, storage and manipulation of a large amount of data from industrial plants. This development has motivated the advancement of research in fault detection, especially based on process history data. Although a large amount of work has been conducted in recent years on the diagnostics of gas turbines, few of them present the use of clustering approaches applied to multivariate time series, adopting PCA similarity factor (SPCA) in order to detect and/or prevent failures. This paper presents a comprehensive method for pattern recognition associated to fault prediction in gas turbines using time series mining techniques. Algorithms comprising appropriate similarity metrics, subsequence matching and fuzzy clustering were applied on data extracted from a Plant Information Management System (PIMS) represented by multivariate time series. A real case study comprising the fault detection in a gas turbine was investigated. The results suggest the existence of a safe way to start the turbine that can be useful to support the development of a dynamic system for monitoring and predicting the probability of failure and for decision-making at operational level.

Abnormal Events Detection in Traffic Data

Detecting abnormal events such as crash is a practical problem that is important to Intelligent Transportation System. By taking advantage of the data recorded by the remote sensors which are deployed along the road, we can perform data mining techniques to see whether there are abnormal events happening on the road. This paper aims at proposing an abnormal-events-detecting method based on the traffic data, which first utilizes outlier detection to generate a fuzzy result set from source data, and then through the time series mining techniques to filter that to obtain an accurate experimental one. Experiment with real-world data shows that our method works satisfactorily in detecting abnormalities such crash, stall and hazard on the road.