CPMAN: Change Point Detection Approach in Time Series Based on the Prediction of Multi-stage Attention Networks

Time series change point detection can identify the locations of abrupt points in many dynamic processes. It can help us to find anomaly data in an early stage. At the same time, detecting change points for long, periodic, and multiple input series data has received a lot of attention recently, and is universally applicable in many fields including power, environment, finance, and medicine. However, the performance of classical methods typically scales poorly for such time series. In this paper, we propose CPMAN, a novel prediction-based change point detection approach via multi-stage attention networks. Our model consists of two key modules. First, in the time series prediction module, we employ the multi-stage attention-based networks and integrate the multi-series fusion mechanism. This module can adaptively extract features from the relevant input series and capture the long-term temporal dependencies. Secondly, in the change point detection module, we use the wavelet analysis-based algorithm to detect change points efficiently and identify the change points and outliers. Extensive experiments are conducted on various real-world datasets and synthetic datasets, proving the superiority and effectiveness of CPMAN. Our approach outperforms the state-of-the-art methods by up to 12.1% on the F1 Score.

Download Full-text

Change Point Enhanced Anomaly Detection for IoT Time Series Data

Water ◽

10.3390/w13121633 ◽

2021 ◽

Vol 13 (12) ◽

pp. 1633

Author(s):

Elena-Simona Apostol ◽

Ciprian-Octavian Truică ◽

Florin Pop ◽

Christian Esposito

Keyword(s):

Time Series ◽

Anomaly Detection ◽

Change Point ◽

Time Series Data ◽

Multivariate Time Series ◽

Change Point Detection ◽

Change Points ◽

Series Data ◽

Prediction And Forecasting ◽

Point Detection

Due to the exponential growth of the Internet of Things networks and the massive amount of time series data collected from these networks, it is essential to apply efficient methods for Big Data analysis in order to extract meaningful information and statistics. Anomaly detection is an important part of time series analysis, improving the quality of further analysis, such as prediction and forecasting. Thus, detecting sudden change points with normal behavior and using them to discriminate between abnormal behavior, i.e., outliers, is a crucial step used to minimize the false positive rate and to build accurate machine learning models for prediction and forecasting. In this paper, we propose a rule-based decision system that enhances anomaly detection in multivariate time series using change point detection. Our architecture uses a pipeline that automatically manages to detect real anomalies and remove the false positives introduced by change points. We employ both traditional and deep learning unsupervised algorithms, in total, five anomaly detection and five change point detection algorithms. Additionally, we propose a new confidence metric based on the support for a time series point to be an anomaly and the support for the same point to be a change point. In our experiments, we use a large real-world dataset containing multivariate time series about water consumption collected from smart meters. As an evaluation metric, we use Mean Absolute Error (MAE). The low MAE values show that the algorithms accurately determine anomalies and change points. The experimental results strengthen our assumption that anomaly detection can be improved by determining and removing change points as well as validates the correctness of our proposed rules in real-world scenarios. Furthermore, the proposed rule-based decision support systems enable users to make informed decisions regarding the status of the water distribution network and perform effectively predictive and proactive maintenance.

Download Full-text

A Novel Change-Point Detection Approach for Monitoring High-Dimensional Traffics in Distributed Systems

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.536-537.499 ◽

2014 ◽

Vol 536-537 ◽

pp. 499-511 ◽

Cited By ~ 1

Author(s):

Li Zhao ◽

Qian Liu ◽

Peng Du ◽

Ge Fu ◽

Wei Cao

Keyword(s):

Time Series ◽

Distributed Systems ◽

Change Point ◽

Principal Component ◽

Change Point Detection ◽

Change Points ◽

Clustering Methods ◽

Detection Approach ◽

Time Series Segmentation ◽

Point Detection

Change-point detection is the problem of finding abrupt changes in time-series. However, the meaningful changes are usually difficult to identify from the original massive traffics, due to high dimension and strong periodicity. In this paper, we propose a novel change-point detection approach, which simultaneously detects change points from all dimensions of the traffics with three steps. We first reduce the dimensions by the classical Principal Component Analysis (PCA), then we apply an extended time-series segmentation method to detect the nontrivial change times, finally we identify the responsible applications for the changes by F-test. We demonstrate through experiments on datasets collected from four distributed systems with 44 applications that the proposed approach can effectively detect the nontrivial change points from the multivariate and periodical traffics. Our approach is more appropriate for mining the nontrivial changes in traffic data comparing with other clustering methods, such as center-based Kmeans and density-based DBSCAN.

Download Full-text

Embedding-based real-time change point detection with application to activity segmentation in smart home time series data

Expert Systems with Applications ◽

10.1016/j.eswa.2021.115641 ◽

2021 ◽

pp. 115641

Author(s):

Unai Bermejo ◽

Aitor Almeida ◽

Aritz Bilbao ◽

Gorka Azkune

Keyword(s):

Time Series ◽

Real Time ◽

Change Point ◽

Smart Home ◽

Time Series Data ◽

Time Change ◽

Change Point Detection ◽

Series Data ◽

Point Detection

Download Full-text

WATCH: Wasserstein Change Point Detection for High-Dimensional Time Series Data

10.1109/bigdata52589.2021.9671962 ◽

2021 ◽

Author(s):

Kamil Faber ◽

Roberto Corizzo ◽

Bartlomiej Sniezynski ◽

Michael Baron ◽

Nathalie Japkowicz

Keyword(s):

Time Series ◽

Change Point ◽

Time Series Data ◽

Change Point Detection ◽

Series Data ◽

High Dimensional ◽

Point Detection

Download Full-text

Multiple change point detection and validation in autoregressive time series data

Statistical Papers ◽

10.1007/s00362-020-01198-w ◽

2020 ◽

Vol 61 (4) ◽

pp. 1507-1528 ◽

Cited By ~ 1

Author(s):

Lijing Ma ◽

Andrew J. Grant ◽

Georgy Sofronov

Keyword(s):

Time Series ◽

Change Point ◽

Time Series Data ◽

Change Point Detection ◽

Series Data ◽

Autoregressive Time Series ◽

Point Detection

Download Full-text

Locally adaptive change-point detection (LACPD) with applications to environmental changes

Stochastic Environmental Research and Risk Assessment ◽

10.1007/s00477-021-02083-0 ◽

2021 ◽

Author(s):

Mehdi Moradi ◽

Manuel Montesino-SanMartin ◽

M. Dolores Ugarte ◽

Ana F. Militino

Keyword(s):

Time Series ◽

Change Point ◽

Error Probability ◽

Type I Error ◽

Change Point Detection ◽

Alternative Methods ◽

Change Points ◽

Type I ◽

Type I Error Probability ◽

Point Detection

AbstractWe propose an adaptive-sliding-window approach (LACPD) for the problem of change-point detection in a set of time-ordered observations. The proposed method is combined with sub-sampling techniques to compensate for the lack of enough data near the time series’ tails. Through a simulation study, we analyse its behaviour in the presence of an early/middle/late change-point in the mean, and compare its performance with some of the frequently used and recently developed change-point detection methods in terms of power, type I error probability, area under the ROC curves (AUC), absolute bias, variance, and root-mean-square error (RMSE). We conclude that LACPD outperforms other methods by maintaining a low type I error probability. Unlike some other methods, the performance of LACPD does not depend on the time index of change-points, and it generally has lower bias than other alternative methods. Moreover, in terms of variance and RMSE, it outperforms other methods when change-points are close to the time series’ tails, whereas it shows a similar (sometimes slightly poorer) performance as other methods when change-points are close to the middle of time series. Finally, we apply our proposal to two sets of real data: the well-known example of annual flow of the Nile river in Awsan, Egypt, from 1871 to 1970, and a novel remote sensing data application consisting of a 34-year time-series of satellite images of the Normalised Difference Vegetation Index in Wadi As-Sirham valley, Saudi Arabia, from 1986 to 2019. We conclude that LACPD shows a good performance in detecting the presence of a change as well as the time and magnitude of change in real conditions.

Download Full-text

Nonparametric Test for Change-Point Detection of IoT Time-Series Data

10.1007/978-3-030-76653-5_5 ◽

2021 ◽

pp. 99-122

Author(s):

Dmitriy Klyushin ◽

Andrii Urazovskyi

Keyword(s):

Time Series ◽

Change Point ◽

Time Series Data ◽

Nonparametric Test ◽

Change Point Detection ◽

Series Data ◽

Point Detection

Download Full-text

Change Point Detection For Clustered Expression Data In Short Time Series.

10.21203/rs.3.rs-605289/v1 ◽

2021 ◽

Author(s):

Miriam Sieg ◽

Lina Katrin Sciesielski ◽

Karin Kirschner ◽

Jochen Kruppa

Keyword(s):

Time Series ◽

Confidence Intervals ◽

Change Point ◽

Mixed Model ◽

Linear Mixed Model ◽

Change Point Detection ◽

Change Points ◽

Point Detection ◽

Over Time ◽

Time Point

Abstract Background: In longitudinal studies, observations are made over time. Hence, the single observations at each time point are dependent, making them a repeated measurement. In this work, we explore a different, counterintuitive setting: At each developmental time point, a lethal observation is performed on the pregnant or nursing mother. Therefore, the single time points are independent. Furthermore, the observation in the offspring at each time point is correlated with each other because each litter consists of several (genetically linked) littermates. In addition, the observed time series is short from a statistical perspective as animal ethics prevent killing more mother mice than absolutely necessary, and murine development is short anyway. We solve these challenges by using multiple contrast tests and visualizing the change point by the use of confidence intervals.Results: We used linear mixed models to model the variability of the mother. The estimates from the linear mixed model are then used in multiple contrast tests.There are a variety of contrasts and intuitively, we would use the Changepoint method. However, it does not deliver satisfying results. Interestingly, we found two other contrasts, both capable of answering different research questions in change point detection: i) Should a single point with change direction be found, or ii) Should the overall progression be determined? The Sequen contrast answers the first, the McDermott the second. Confidence intervals deliver effect estimates for the strength of the potential change point. Therefore, the scientist can define a biologically relevant limit of change depending on the research question.Conclusion: We present a solution with effect estimates for short independent time series with observations nested at a given time point. Multiple contrast tests produce confidence intervals, which allow determining the position of change points or to visualize the expression course over time. We suggest to use McDermott’s method to determine if there is an overall significant change within the time frame, while Sequen is better in determining specific change points. In addition, we offer a short formula for the estimation of the maximal length of the time series.

Download Full-text

Using Graph Spectral to solve Change Point Detection Problems

10.5753/eniac.2018.4461 ◽

2018 ◽

Author(s):

Luis Gustavo C. Uzai ◽

André Y. Kashiwabara

Keyword(s):

Time Series ◽

Change Point ◽

State Of The Art ◽

The State ◽

Change Point Detection ◽

Change Points ◽

Graph Spectrum ◽

Point Detection ◽

Over Time

Time series are sequence of values distributed over time. Analyzing time series is important in many areas including medical, financial, aerospace, commercial and entertainment. Change Point Detection is the problem of identifying changes in meaning or distribution of data in a time series. This article presents Spec, a new algorithm that uses the graph spectrum to detect change points. The Spec was evaluated using the UCR Archive which is a large da- tabase of different time series. Spec performance was compared to the PELT, ECP, EDM, and gSeg algorithms. The results showed that Spec achieved a better accuracy compared to the state of the art in some specific scenarios and as efficient as in most cases evaluated.

Download Full-text