Time Series Data Mining Techniques Research Articles

A shared multi-scale lightweight convolution generative network for few-shot multivariate time series forecasting

Time series forecasting is an important time series data mining technique. Among them, multivariate time series (MTS) forecasting has received extensive attention in many fields. However, many existing MTS forecasting models usually rely on a large amount of labeled data for model training, and data collection and labeling are difficult in real systems. The insufficient amount of data makes it difficult for the model to fully learn the intrinsic patterns and features of the data, which not only increases the prediction error, but also makes it hard to obtain satisfactory prediction results. To address this challenge, we propose a shared multi-scale lightweight convolution generative (SMLCG) network for few-shot multivariate time series forecasting by using samples generation strategy. The overall goal is to design a shared multi-scale feature generation prediction framework that generates data highly similar to the original sample and enriches the training sample to improve prediction accuracy. Specifically, the MTS is divided into different scales, and the multi-scale feature fusion module is utilized to capture and fuse the MTS information in different spatial dimensions to eliminate the heterogeneity among the data. Then, the key information in the multi-scale features is captured by a lightweight convolution generative network, and the feature weights are dynamically assigned to explore the change information. In addition, a spatio-temporal memory module is designed based on the parameter sharing strategy to capture the spatio-temporal dynamic relationship of sequences by learning the common knowledge in multi-scale features, thus improving the robustness and generalization ability. Through comprehensive experiments on four publicly available datasets and comparisons with other reported models, it is demonstrated that the SMLCG model can efficiently generate approximate samples in the few-shot case and provide excellent prediction results. The architecture of SMLCG serves as a valuable reference for practical solutions to address the few-shot problem in multivariate time series.

Multi-Querying: A Subsequence Matching Approach to Support Multiple Queries

The widespread use of sensors has resulted in an unprecedented amount of time series data. Time series mining has experienced a particular surge of interest, among which, subsequence matching is one of the most primary problem that serves as a foundation for many time series data mining techniques, such as anomaly detection and classification. In literature there exist many works to study this problem. However, in many real applications, it is uneasy for users to accurately and clearly elaborate the query intuition with a single query sequence. Consequently, in this paper, we address this issue by allowing users to submit a small query set, instead of a single query. The multiple queries can embody the query intuition better. In particular, we first propose a novel probability-based representation of the query set. A common segmentation is generated which can approximate the queries well, in which each segment is described by some features. For each feature, the corresponding values of multiple queries are represented as a Gaussian distribution. Then, based on the representation, we design a novel distance function to measure the similarity of one subsequence to the multiple queries. Also, we propose a breadth-first search strategy to find out similar subsequences. We have conducted extensive experiments on both synthetic and real datasets, and the results verify the superiority of our approach.

Time Series Data Mining for Sport Data: a Review

Abstract Time series data mining deals with extracting useful and meaningful information from time series data. Recently, the increasing use of temporal data, in particular time series data, has received much attention in the literature. Since most of sports data contain time information, it is natural to consider the temporal dimension in form of time series. However, in sports, the effective use of time series data mining techniques is still under development. The main goal of this paper is therefore to serve as an introduction to time series data mining and a glossary for interested researchers from the sports community. The paper gives an overview about current data mining tasks and tries to identify their potential research direction for further investigation. Furthermore, we want to draw more attention with respect to the importance of mining approaches with sport data and their particular challenges beyond usual time series data mining tasks.

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.

Medium-term forecasts for salinity rates and groundwater levels

An increase in demand for groundwater associated with uncontrolled consumption has led to the depletion of groundwater wells. As a result, the mixing of seawater with groundwater increases the salinity rates, especially in areas where wells are close to the Mediterranean Sea in Gaza. In this paper, we use time-series data mining techniques to forecast the salinity rates and levels of groundwater in Deir El-Balah city—Gaza Strip. For this purpose, five forecasting techniques were applied on two data sets gained from the Palestinian Water Authority of Deir El-Balah City: salinity rates and groundwater levels (GL). The used forecasting algorithms are: exponential smoothing (ETS), state space model with Box–Cox transformation, ARMA errors, trend and seasonal components (TBATS), auto-regressive integrated moving average (ARIMA), ARIMA combined with: neural network (NN), ETS, and TBATS model. The best performance of the applied algorithms of salinity data according to Mean Absolute Percentage Error (MAPE) measure on the well S-69 was: ARIMA (MAPE = 4.2%), and (ARIMA + TBATS) for K-21 and K-20 which gave the MAPE = 5.4% and 4.0%, respectively. On the other hand, ARIMA was the most convenient algorithm to forecast the salinity rates of GL for S-50 (MAPE = 2.5) and ARIMA + NN (MAPE = 2.1) for J-103. The results demonstrated that in the period (2018–2023), the salinity rates will increase for S-69, K-20, and K-21 in comparison with the period (2012–2017) by 7.1%, 69.6%, and 55.7%, respectively. While the water levels of the wells: S-50 and J-103 will be lower than the sea water level in the period (2018–2023) by 1.89% and 6.37% in comparison with the period (2012–2017).

Characterizing Land Surface Phenology and Exotic Annual Grasses in Dryland Ecosystems Using Landsat and Sentinel-2 Data in Harmony

Invasive annual grasses, such as cheatgrass (Bromus tectorum L.), have proliferated in dryland ecosystems of the western United States, promoting increased fire activity and reduced biodiversity that can be detrimental to socio-environmental systems. Monitoring exotic annual grass cover and dynamics over large areas requires the use of remote sensing that can support early detection and rapid response initiatives. However, few studies have leveraged remote sensing technologies and computing frameworks capable of providing rangeland managers with maps of exotic annual grass cover at relatively high spatiotemporal resolutions and near real-time latencies. Here, we developed a system for automated mapping of invasive annual grass (%) cover using in situ observations, harmonized Landsat and Sentinel-2 (HLS) data, maps of biophysical variables, and machine learning techniques. A robust and automated cloud, cloud shadow, water, and snow/ice masking procedure (mean overall accuracy >81%) was implemented using time-series outlier detection and data mining techniques prior to spatiotemporal interpolation of HLS data via regression tree models (r = 0.94; mean absolute error (MAE) = 0.02). Weekly, cloud-free normalized difference vegetation index (NDVI) image composites (2016–2018) were used to construct a suite of spectral and phenological metrics (e.g., start and end of season dates), consistent with information derived from Moderate Resolution Image Spectroradiometer (MODIS) data. These metrics were incorporated into a data mining framework that accurately (r = 0.83; MAE = 11) modeled and mapped exotic annual grass (%) cover throughout dryland ecosystems in the western United States at a native, 30-m spatial resolution. Our results show that inclusion of weekly HLS time-series data and derived indicators improves our ability to map exotic annual grass cover, as compared to distribution models that use MODIS products or monthly, seasonal, or annual HLS composites as primary inputs. This research fills a critical gap in our ability to effectively assess, manage, and monitor drylands by providing a framework that allows for an accurate and timely depiction of land surface phenology and exotic annual grass cover at spatial and temporal resolutions that are meaningful to local resource managers.

Using methods of time series data mining to recognize the influences of environmental factors on bullous pemphigoid

ABSTRACTBullous pemphigoid (BP) is an acute or chronic autoimmune skin disease. It is reported that poor conditions of general health and increased mortality rates are observed in BP patients, although there is a paucity of studies on the pathophysiologic reasons for the phenomena. In the study, we applied various time series data mining techniques to validate the effect of environmental factors on BP. Several statistical methods were used to explore the seasonal incidence patterns of BP, and validate the significance of the obtained patterns. The correlation measure was employed to evaluate the relationship between meteorological variables and BP incidence. The Fourier–Gaussian Decomposition (FGD) algorithm developed by our research group is a significant contribution. The FGD was also adopted to recognize the seasonal incidence distributions of BP. The study indicated that the monthly incident counts for BP significantly increased in August. The correlation measure verified strong correlation of BP incidence with high ambient temperature. Furthermore, with respect to its ability to extract hidden patterns from input sequences, the effectiveness of the FGD algorithm was compared and validated with conventional methods for time series data mining. Hence, it is expected that the mined epidemiological properties may aid in disease prevention of BP.

Multivariate Sensor Data Analysis for Oil Refineries and Multi-mode Identification of System Behavior in Real-time

Large-scale oil refineries are equipped with mission-critical heavy machinery (boilers, engines, turbines, and so on) and are continuously monitored by thousands of sensors for process efficiency, environmental safety, and predictive maintenance purposes. However, sensors themselves are also prone to errors and failure. The quality of data received from these sensors should be verified before being used in system modeling. There is a need for reliable methods and systems that can provide data validation and reconciliation in real-time with high accuracy. In this paper, we develop a novel method for real-time data validation, gross error detection and classification over multivariate sensor data streams. The validated and high-quality data obtained from these processes is used for pattern analysis and modeling of industrial plants. We obtain sensor data from the power and petrochemical plants of an oil refinery and analyze them using various time-series modeling and data mining techniques that we integrate into a complex event processing engine. Next, we study the computational performance implications of the proposed methods and uncover regimes where they are sustainable over fast streams of sensor data. Finally, we detect shifts among steady-states of data, which represent systems’ multiple operating modes and identify the time when a model reconstruction is required using DBSCAN clustering algorithm.

A Comprehensive Survey of Data Mining Techniques on Time Series Data for Rainfall Prediction

Time series data available in huge amounts can be used in decision-making. Such time series data can be converted into information to be used for forecasting. Various techniques are available for prediction and forecasting on the basis of time series data. Presently, the use of data mining techniques for this purpose is increasing day by day. In the present study, a comprehensive survey of data mining approaches and statistical techniques for rainfall prediction on time series data was conducted. A detailed comparison of different relevant techniques was also conducted and some plausible solutions are suggested for efficient time series data mining techniques for future algorithms.Â

A soft computing framework for classifying time series based on fuzzy sets of events

Time series are sequences of data gathered over a period of time that emerge in different domains and whose analysis requires the application of specialized techniques, like, for example, data mining. Many existing time series data mining techniques, like the discrete Fourier transform (DFT), offer solutions for analysing whole time series. Often, however, it is more important to analyse certain regions of interest, known as events, rather than whole time series. Event identification is a highly complex task, as it is not always possible to determine with absolute certainty whether or not a segment of a time series is an event. In such cases, the best practice is to establish the certainty of this segment being a time series event, thus outputting a fuzzy set of events.In this paper we propose a framework that is capable of identifying events and establishing the degree of certainty that a domain expert would assign to the identified events based on a previous training process assisted by a panel of experts. Having identified the events, the proposed framework can be used to classify time series. This is done by means of a process that combines time series comparison and time series reference model generation by analysing the events contained in the respective time series and the certainties of the identified events. The proposed framework is an evolution of an earlier framework that we developed which did not apply soft computing techniques to identify and manage the time series events.We have used our framework to classify times series generated in the electroencephalography (EEG) area. EEG is a neurological exploration used to diagnose nervous system disorders. The performance of the framework was evaluated in terms of classification accuracy. The results confirmed that, thanks to the use of soft computing techniques, the new framework substantially improves the time series classification results of its predecessor.

Temporal knowledge discovery in big BAS data for building energy management

With the advances of information technologies, today's building automation systems (BASs) are capable of managing building operational performance in an efficient and convenient way. Meanwhile, the amount of real-time monitoring and control data in BASs grows continually in the building lifecycle, which stimulates an intense demand for powerful big data analysis tools in BASs. Existing big data analytics adopted in the building automation industry focus on mining cross-sectional relationships, whereas the temporal relationships, i.e., the relationships over time, are usually overlooked. However, building operations are typically dynamic and BAS data are essentially multivariate time series data. This paper presents a time series data mining methodology for temporal knowledge discovery in big BAS data. A number of time series data mining techniques are explored and carefully assembled, including the Symbolic Aggregate approXimation (SAX), motif discovery, and temporal association rule mining. This study also develops two methods for the efficient post-processing of knowledge discovered. The methodology has been applied to analyze the BAS data retrieved from a real building. The temporal knowledge discovered is valuable to identify dynamics, patterns and anomalies in building operations, derive temporal association rules within and between subsystems, assess building system performance and spot opportunities in energy conservation.

Based on ETEO Pattern Abnormal Event Detection in Wireless Sensor Networks

This paper presents a new algorithm for the detection of abnormal events in Wireless Sensor Networks (WSN). Abnormal events are sets of data points that correspond to interesting patterns in the underlying phenomenon that the network monitors. This algorithm is inspired from time-series data mining techniques and transforms a stream of sensor readings into an Extension Temporal Edge Operator (ETEO) of time series pattern representation, and then extracts the three eigenvalue of each sub-pattern, that is, patterns length, patterns slope and patterns mean to map time series to feature space, and finally uses local outlier factor to detect abnormal pattern in this feature space. Experiments on synthetic and real data show that the definition of pattern outlier is reasonable and this algorithm is efficient to detect outliers in WSN.

Discovering Flood Recession Pattern in Hydrological Time Series Data Mining during the Post Monsoon Period

his paper examines the flood recession pattern for the river discharge data in the river Brahmaputra basin. The months from October to December comes under the post monsoon season. In this paper, with the help of time series data mining techniques, the analysis has made for hydrological daily discharge time series data, measured at the Panchratna station during the post monsoon in the river Brahmaputra under Brahmaputra and Barak Basin Organization after the high flood. Statistical analysis has made for standardization of data. K-means clustering, Dynamic Time Warping(DTW), Agglomerative Hierarchical Clustering(AHC) and Ward's criterion are used to cluster and discover the discharge patterns in terms of the autoregressive model. A forecast model has been developed for the discharge process. For validation of the recession pattern, Gauge-Discharge Curve, Water Label Hydrographs, Rainfall Bar Graphs have been developed and also discharge recession coefficient has been calculated. This study gives the behavioral characteristics of rivers discharge during recession of high floods with the time series data mining.

Time-Series Data Mining in Transportation: A Case Study on Singapore Public Train Commuter Travel Patterns

The adoption of smart cards technologies and automated data collection systems (ADCS) in transportation domain had provided public transport planners opportunities to amass a huge and continuously increasing amount of time-series data about the behaviors and travel patterns of commuters. However the explosive growth of temporal related databases has far outpaced the transport planners’ ability to interpret these data using conventional statistical techniques, creating an urgent need for new techniques to support the analyst in transforming the data into actionable information and knowledge. This research study thus explores and discusses the potential use of time-series data mining, a relatively new framework by integrating conventional time-series analysis and data mining techniques, to discover actionable insights and knowledge from the transportation temporal data. A case study on the Singapore public train transit will also be used to demonstrate the time-series data-mining framework and methodology.

Automatic Detection and Classification of Unsafe Events During Power Wheelchair Use.

Using a powered wheelchair (PW) is a complex task requiring advanced perceptual and motor control skills. Unfortunately, PW incidents and accidents are not uncommon and their consequences can be serious. The objective of this paper is to develop technological tools that can be used to characterize a wheelchair user’s driving behavior under various settings. In the experiments conducted, PWs are outfitted with a datalogging platform that records, in real-time, the 3-D acceleration of the PW. Data collection was conducted over 35 different activities, designed to capture a spectrum of PW driving events performed at different speeds (collisions with fixed or moving objects, rolling on incline plane, and rolling across multiple types obstacles). The data was processed using time-series analysis and data mining techniques, to automatically detect and identify the different events. We compared the classification accuracy using four different types of time-series features: 1) time-delay embeddings; 2) time-domain characterization; 3) frequency-domain features; and 4) wavelet transforms. In the analysis, we compared the classification accuracy obtained when distinguishing between safe and unsafe events during each of the 35 different activities. For the purposes of this study, unsafe events were defined as activities containing collisions against objects at different speed, and the remainder were defined as safe events. We were able to accurately detect 98% of unsafe events, with a low (12%) false positive rate, using only five examples of each activity. This proof-of-concept study shows that the proposed approach has the potential of capturing, based on limited input from embedded sensors, contextual information on PW use, and of automatically characterizing a user’s PW driving behavior.

A Comparative Study on FFNN and ARIMA Model in the Presence of Outliers

Time series data mining (TSDM) techniques explores large amount of time series data in search of interesting relationships among variables. The TSDM methods overcome limitations including stationarity and linearity requirements of traditional time series analysis by adapting data mining concepts for analyzing time series data. The Feed Forward Neural Net is one of the most widely used neural nets. In this paper, the Feed Forward Neural Nets architecture is examined and compared with Statistical Time Series Auto regressive integrated moving average (ARIMA) model for prediction of agricultural production. The performance by ANN model and Time series model for prediction are examined using visualization technique and statistical test and the results are illustrated numerically and graphically

Perception-based approach to time series data mining

Time series data mining (TSDM) techniques permit exploring large amounts of time series data in search of consistent patterns and/or interesting relationships between variables. TSDM is becoming increasingly important as a knowledge management tool where it is expected to reveal knowledge structures that can guide decision making in conditions of limited certainty. Human decision making in problems related with analysis of time series databases is usually based on perceptions like “end of the day”, “high temperature”, “quickly increasing”, “possible”, etc. Though many effective algorithms of TSDM have been developed, the integration of TSDM algorithms with human decision making procedures is still an open problem. In this paper, we consider architecture of perception-based decision making system in time series databases domains integrating perception-based TSDM, computing with words and perceptions, and expert knowledge. The new tasks which should be solved by the perception-based TSDM methods to enable their integration in such systems are discussed. These tasks include: precisiation of perceptions, shape pattern identification, and pattern retranslation. We show how different methods developed so far in TSDM for manipulation of perception-based information can be used for development of a fuzzy perception-based TSDM approach. This approach is grounded in computing with words and perceptions permitting to formalize human perception-based inference mechanisms. The discussion is illustrated by examples from economics, finance, meteorology, medicine, etc.

Forecasting the volatility of stock price index

Accurate volatility forecasting is the core task in the risk management in which various portfolios’ pricing, hedging, and option strategies are exercised. Prior studies on stock market have primarily focused on estimation of stock price index by using financial time series models and data mining techniques. This paper proposes hybrid models with neural network and time series models for forecasting the volatility of stock price index in two view points: deviation and direction. It demonstrates the utility of the hybrid model for volatility forecasting. This model demonstrates the utility of the neural network forecasting combined with time series analysis for the financial goods.

Diagnostics of eccentricities and bar/end-ring connector breakages in polyphase induction motors through a combination of time-series data mining and time-stepping coupled fe~state-space techniques

This paper develops the foundations of a technique for detection and categorization of dynamic/static eccentricities and bar/end-ring connector breakages in squirrel-cage induction motors that is not based on the traditional Fourier transform frequency-domain spectral analysis concepts. Hence, this approach can distinguish between the of each of the following faults: eccentricities, broken bars, and broken end-ring connectors in such induction motors. Furthermore, the techniques presented here can extensively and economically predict and characterize faults from the induction machine adjustable-speed drive design data without the need to have had actual fault data from field experience. This is done through the development of dual-track studies of fault simulations and, hence, simulated fault signature data. These studies are performed using our proven time-stepping coupled finite-element-state-space method to generate fault case performance data, which contain phase current waveforms and time-domain torque profiles. Then, from this data, the fault cases are classified by their inherent characteristics, so-called or fingerprints. These fault signatures are extracted or mined here from the fault case data using our novel time-series data mining technique. The dual track of generating fault data and mining fault signatures was tested here on dynamic and static eccentricities of 10% and 30% of air-gap height as well as cases of one, three, six, and nine broken bars and three, six, and nine broken end-ring connectors. These cases were studied for proof of principle in a 208 V 60 Hz four-pole 1.2 hp squirrel-cage three-phase induction motor. The paper presents faulty and healthy performance characteristics and their corresponding so-called phase space diagnoses that show distinct fault signatures of each of the above-mentioned motor faults.

Diagnostics of bar and end-ring connector breakage faults in polyphase induction motors through a novel dual track of time-series data mining and time-stepping coupled FE-state space modeling

This paper develops the fundamental foundations of a technique for detection of faults in induction motors that is not based on the traditional Fourier transform frequency domain approach. The technique can extensively and economically characterize and predict faults from the induction machine adjustable speed drive design data. This is done through the development of dual-track proof-of-principle studies of fault simulation and identification. These studies are performed using our proven time stepping coupled finite element-state space method to generate fault case data. Then, the fault cases are classified by their inherent characteristics, so-called "signatures" or "fingerprints." These fault signatures are extracted or mined here from the fault case data using our novel time series data mining technique. The dual-track of generating fault data and mining fault signatures was tested here on three, six, and nine broken bar and broken end-ring connectors in a 208-volt, 60-Hz, 4-pole, 1.2-hp, squirrel cage 3-phase induction motor.