Change Point Detection Algorithm Research Articles

Online detection of offsets in GPS time series

This paper deals with the online offset detection in GPS time series recorded in volcanic areas. The interest for this problem lies in the fact that an offset can indicate the opening of eruptive fissures. A Change Point Detection algorithm is applied to carry out, in an online framework, the offset detection. Experimental results show that the algorithm is able to recognize the offset generated by the Mount Etna eruption, occurred on December 24, 2018, with a delay of about 4 samples, corresponding to 40 min, compared to the best offline detection. Furthermore, analysis of the trade-off between success and false alarms is carried out and discussed.

Request–Response and Censoring-Based Energy-Efficient Decentralized Change-Point Detection With IoT Applications

Change-point detection (CPD) from streaming data is a fundamental problem in statistical signal processing and has wide applications in wireless sensor networks and the Internet of Things (IoT), such as environmental monitoring and physical activity detection. In various detection schemes, decentralized detection without fusion center is becoming increasingly popular for IoT applications since it enjoys the benefit of strong robustness and security. However, it faces the big challenge of high energy consumption. In this work, we propose an energy-efficient decentralized CPD algorithm called request–response and censoring-based cumulative sum. Specifically, we design a novel communication strategy based on request–response and censoring scheme, which could help the sensors extract useful information from the neighbor sets with a low amount of communication. Furthermore, to provide a guideline on selecting the parameters involved in our algorithm, we theoretically analyze the performance of the proposed algorithm under a simplification for the most interesting cases, in terms of two standard criteria average running length (ARL) and expected detection delay (EDD). Numerical simulations are conducted to verify the effectiveness of the proposed algorithm. Moreover, we test the feasibility of the proposed algorithm in the physical activity CPD task. The experimental results on the real-world data set demonstrate its high energy efficiency and small detection delay.

Technical Appendix for 'Exploring Breaks in the Distribution of Stock Returns: Empirical Evidence from Apple Inc.'

This Appendix contains technical details on the changepoint detection algorithms applied in the paper.

A Novel Online Change Point Detection Using an Approximate Random Blanket and the Line Process Energy

In this paper, we focus on the problem of change point detection in piecewise constant signals. This problem is central to several applications such as human activity analysis, speech or image analysis and anomaly detection in genetics. We present a novel window-sliding algorithm for an online change point detection. The proposed approach considers a local blanket of a global Markov Random Field (MRF) representing the signal and its noisy observation. For each window, we define and solve the local energy minimization problem to deduce the gradient on each edge of the MRF graph. The gradient is then processed by an activation function to filter the weak features and produce the final jumps. We demonstrate the effectiveness of our method by comparing its running time and several detection metrics with state of the art algorithms.

Network change point detection in resting-state functional connectivity dynamics of mild cognitive impairment patients

Background/Objective: This study aims to characterize the differences on the short-term temporal network dynamics of the undirected and weighted whole-brain functional connectivity between healthy aging individuals and people with mild cognitive impairment (MCI). The Network Change Point Detection algorithm was applied to identify the significant change points in the resting-state fMRI register, and we analyzed the fluctuations in the topological properties of the sub-networks between significant change points. Method: Ten MCI patients matched by gender and age in 1:1 ratio to healthy controls screened during patient recruitment. A neuropsychological evaluation was done to both groups as well as functional magnetic images were obtained with a Philips 3.0T. All the images were preprocessed and statistically analyzed through dynamic point estimation tools. Results: No statistically significant differences were found between groups in the number of significant change points in the functional connectivity networks. However, an interaction effect of age and state was detected on the intra-participant variability of the network strength. Conclusions: The progression of states was associated to higher variability in the patient's group. Additionally, higher performance in the prospective and retrospective memory scale was associated with higher median network strength.

An assessment of practitioners approaches to forecasting in the presence of changepoints

AbstractA common challenge in time series is to forecast data that suffer from structural breaks or changepoints which complicate modeling. If we naively forecast using one model for the whole data, the model will be incorrect, and thus, our forecast error will be large. There are two common practices to account for these changepoints when the goal is forecasting: (1) preprocess the data to identify the changepoints, incorporating them as dummy variables in modeling the whole data, and (2) include the changepoint estimation into the model and forecast using the model fit to the last segment. This article examines these two practices, using the computationally exact Pruned Exact Linear Time (PELT) algorithm for changepoint detection, comparing and contrasting them in the context of an important Software Engineering application.

Time series analysis of COVID-19 infection curve: A change-point perspective

In this paper, we model the trajectory of the cumulative confirmed cases and deaths of COVID-19 (in log scale) via a piecewise linear trend model. The model naturally captures the phase transitions of the epidemic growth rate via change-points and further enjoys great interpretability due to its semiparametric nature. On the methodological front, we advance the nascent self-normalization (SN) technique (Shao, 2010) to testing and estimation of a single change-point in the linear trend of a nonstationary time series. We further combine the SN-based change-point test with the NOT algorithm (Baranowski et al., 2019) to achieve multiple change-point estimation. Using the proposed method, we analyze the trajectory of the cumulative COVID-19 cases and deaths for 30 major countries and discover interesting patterns with potentially relevant implications for effectiveness of the pandemic responses by different countries. Furthermore, based on the change-point detection algorithm and a flexible extrapolation function, we design a simple two-stage forecasting scheme for COVID-19 and demonstrate its promising performance in predicting cumulative deaths in the U.S.

A cyber-physical environment for detecting exceptional and dangerous human behavior in the home by sensors and its verification by computer simulation

Today aging of society is becoming increasingly important. Many people want to live at home as long as possible. Very often this leads to accidents at home, which are unnoticed for a long time and therefore cause severe complications. We propose a cyber-physical system consisting of sensors for motion, light, temperature, and so on, which monitors the behavior of elderly people but is non-invasive in the sense that the persons are not observed by a camera and must not wear certain sensors on their body. Unusual behavior or accidents are registered in-time by a change point detection algorithm based on Markov chains, which does not store any data with the exception of the last datum, so that also the integrity of the elderly people is preserved. For verification of this algorithm, a cyber-physical test environment has been programmed which allows to simulate human common and uncommon behavior in house during day and night. This simulation environment is based on behavior trees and decision theory. The verification results suggest that the implemented cyber-physical system for change point and anomaly detection could be successfully used in the real environment of elderly people to give them help in-time when an accident occurs.

Dynamic Segmentation Of Youth Accelerometer Data By Sojourn And Change Point Detection Methods

Dynamic segmentation algorithms are used to find activity transitions in accelerometer data. Youth Sojourn models use a crude algorithm, which may be improved by instead using a change point detection (CPD) algorithm. Pruned exact linear time (PELT) is a CPD algorithm that finds transitions by minimizing a cost function while iterating over the data and pruning out inviable transition points. PURPOSE: To compare the performance of youth Sojourn and PELT. METHODS: Raw acceleration data (hip-worn ActiGraph GT9X) from 86 youth (age 6-18 yrs; 48% male; 16% overweight/obese) were processed using Sojourn and PELT. Participants performed two semi-structured activity routines on separate days, with each visit lasting approximately 2-2.5 h. A total of 16 activities (eight each day) were performed, twice each, and the study protocol was designed to promote variability in the ordering and duration of activities. Throughout each trial, direct observation was performed using focal sampling, which served as a criterion measure of when activity transitions occurred. Sojourn and PELT were compared to the criterion using the transition pairing method, with a maximum of 5-s lag time allowed for a prediction to be considered a true positive. Performance metrics were recall, precision, and root mean squared error (RMSE). The metrics were calculated for each participant (both visits combined), after which paired t-tests were used to compare Sojourn-vs-PELT means for each metric. RESULTS: Values are mean ± SD. Recall was similar for Sojourn (49.6% ± 9.0%) and PELT (51.5% ± 9.2%, p = 0.15), and the same was true for RMSE (2.9 ± 0.3 s for Sojourn, versus 3.1 ± 0.4 s for PELT, p < 0.001). However, precision for Sojourn (21.7% ± 4.9%) was substantially lower than for PELT (38.7% ± 11.0%, p < 0.001). CONCLUSION: Youth Sojourn models may benefit from replacing their current segmentation algorithms with CPD algorithms like PELT. Thus, CPD warrants further investigation. Supported by NIH R01HD083431

Resting-State Functional Connectivity Dynamics in Healthy Aging: An Approach Through Network Change Point Detection.

This study aims at assessing the impact of age on the short-term temporal dynamics of the topological properties of the undirected and weighted whole-brain functional connectivity (FC) networks. We studied the association between the participant's age and the number of significant change points detected through the Network Change Point Detection algorithm. Secondary, we defined state as the resting-state functional magnetic resonance imaging (rs-fMRI) subsequence between two significant change points, obtaining the FC network in each state and participant and characterized their network topological properties. The data comprise the rs-fMRI sequences of 114 healthy individuals combined from 3 different studies conducted at the Department of Medicine, School of Medicine and Health Sciences, University of Barcelona. Participants were healthy people in the absence of any pathology that could interfere with the scanning procedures, as well as any chronic illness that implied a short-lived situation. Topological properties of everyone's FC networks were characterized by their network strength, transitivity, characteristic path length, and small-worldness, analyzing the effect of age in those observed distributions. To that effect, we constructed a mixed linear model for each network topological property with age, state, and state duration as the linear predictors. Several statistically significant relationships have been estimated between the indicators of the FC networks that show a certain regular pattern of change in the networks during the time of registration at the resting fMRI paradigm. These dynamic changes seem to be related to the age of each group studied. Healthy aging could be characterized by FC dynamics patterns.

Human Activity Prediction in Smart Home Time Series Data by using Change Point Detection

A Sensor is attached to a house which send message now and then about the housemates or the old people activities. Whether they had taken food in the correct time or not and other activities they perform in the proper time or not. Change Point Detection (CPD) is a matter of perspective which discovers deviating from what are normal or usual changes within the housemates. Any abnormal changes in the housemates have been identifying the presence of time points. The dissimilar changes occurs called SEPERATION Change Point Detection. It will not coincide when a remarkable occurrence of events at any points. Change Point Detection (CPD) occurs at the same time and in the problem of finding unexpected changes in facts and statistics collected together for references and analysis and in the property of the time series changes. An unusual real-time not involving any assumptions as to the form or in the parameters of a frequency distribution change point detection algorithm called Separation , It is used to calculate as a parting to recognize change points in fully measurement characteristics relating to measurements having sufficient depth and substance to be in time series. To ameliorate the order of this algorithm used in ARIMA with SEP algorithm. ARIMA model is used for predicting the Time series forecasting result. If emergency is occur then automatically send notification to caring person. The proposed work can decreasing computational cost and also improves the detection accuracy in the quality or fact of being useful of proposed technique.

An FPGA multiprocessor architecture for Bayesian online change point detection using stochastic computation

In this paper we report on an event-based stochastic architecture for the Adams/McKay Bayesian Online Change Point Detection algorithm (BOCPD) [1]. In the stochastic computational structures, probabilities are represented natively as stochastic events and computation is carried out directly with these probabilities and not probability density functions. A fully programmable BOCPD processor is synthesized in VHDL. The BOCPD algorithm with on-line learning, to perform foreground/background image segmentation with online learning. Running on a single Kintex 7 FPGA (Opal Kelly XEM7350-K410T) the architecture is capable of real-time processing a 160 × 120 pixels image, at 10 frames per second.

A segmented generalized Markov regime-switching model with its application in financial time series data

ABSTRACTMovements of equity indices are very important information for an investment decision. Empirical studies illustrate that the movements switch among different regimes. The Markov regime-switching model has important applications to such analysis. However, parameters estimated under normality assumption might not be stable and the corresponding change-point detection algorithm might face some challenges when either the error distribution is heavy-tailed or observed data contain outliers. In this paper, we relax the normality assumption and propose a generalized Markov regime-switching (GMRS) model. We propose a GMRS model based change-point detection algorithm, which is tested on both simulation data and Hang Seng monthly index. Simulation studies show that this algorithm can improve the accuracy of identifying change-points when either the error distribution is heavy-tailed or observed data contain outliers. It is also evident that the identified change-points on Hang Seng monthly index data match the observed market behaviours.

On Matched Filtering for Statistical Change Point Detection

Non-parametric and distribution-free two-sample tests have been the foundation of many change point detection algorithms. However, randomness in the test statistic as a function of time makes them susceptible to false positives and localization ambiguity. We address these issues by deriving and applying filters matched to the expected temporal signatures of a change for various sliding window, two-sample tests under IID assumptions on the data. These filters are derived asymptotically with respect to the window size for the Wasserstein quantile test, the Wasserstein-1 distance test, Maximum Mean Discrepancy squared (MMD <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ), and the Kolmogorov-Smirnov (KS) test. The matched filters are shown to have two important properties. First, they are distribution-free, and thus can be applied without prior knowledge of the underlying data distributions. Second, they are peak-preserving, which allows the filtered signal produced by our methods to maintain expected statistical significance. Through experiments on synthetic data as well as activity recognition benchmarks, we demonstrate the utility of this approach for mitigating false positives and improving the test precision. Our method allows for the localization of change points without the use of ad-hoc post-processing to remove redundant detections common to current methods. We further highlight the performance of statistical tests based on the Quantile-Quantile (Q-Q) function and show how the invariance property of the Q-Q function to order-preserving transformations allows these tests to detect change points of different scales with a single threshold within the same dataset.

A Distributed Intrusion Detection Scheme for Cloud Computing

Intrusion detection systems (IDS) is an important security measure used to secure cloud resources, however, IDS often suffer from poor detection accuracy due to coordinated attacks such as a DDoS. Various research on distributed IDSs have been proposed to detect DDoS however, the limitations of these works the lack of technique to determine an appropriate period to share attack information among nodes in the distributed IDS. Therefore, this article proposes a distributed IDS that uses a binary segmentation change point detection algorithm to address the appropriate period to send attack information to nodes in distributed IDS and using parallel Stochastic Gradient Descent with Support Vector Machine (SGD-SVM) to achieve the distributed detection. The result of the proposed scheme was implemented in Apache Spark using NSL-KDD benchmark intrusion detection dataset. Experimental results show that the proposed distributed intrusion detection scheme outperforms existing distributed IDS for cloud computing.

On study for change point regression problems using a difference-based regression model

This paper derive a method to solve change point regression problems via a process for obtaining consequential results using properties of a difference-based intercept estimator first introduced by Park and Kim (Communications in Statistics - Theory Methods, 2019) for outlier detection in multiple linear regression models. We describe the statistical properties of the difference-based regression model in a piecewise simple linear regression model and then propose an efficient algorithm for change point detection. We illustrate the merits of our proposed method in the light of comparison with several existing methods under simulation studies and real data analysis. This methodology is quite valuable, “no matter what regression lines” and “no matter what the number of change points”.

A shape-based cutting and clustering algorithm for multiple change-point detection

The detection of multiple change-points without prior knowledge of the number and location of the change-points is investigated in this paper. By making full use of the geometric information of local statistics, a new change-point detection algorithm called the shape-based cutting and clustering (SCC) algorithm is established. There are three key techniques in the proposed SCC procedure: data-driven threshold, adaptive bandwidth and single peak recognition. Our simulation results show that the proposed method is highly competitive in terms of computational speed and effectiveness. In order to validate the feasibility of the proposed algorithm, we apply the methodology to an operational problem in renewable integrated electrical distribution networks. The results of the real data analysis illustrate the effectiveness of the proposed algorithm.

Count-based change point detection via multi-output log-Gaussian Cox processes

The ability to detect change points is a core skill in system monitoring and prognostics. When data take the form of frequencies, i.e., count data, counting processes such as Poisson processes are extensively used for modeling. However, many existing count-based approaches rely on parametric models or deterministic frameworks, failing to consider complex system uncertainty based on temporal and environmental contexts. Another challenge is analyzing interrelated events simultaneously to detect change points that can be missed by independent analyses. This article presents a Multi-Output Log-Gaussian Cox Process with a Cross-Spectral Mixture kernel (MOLGCP-CSM) as a count-based change point detection algorithm. The proposed model employs MOLGCP to flexibly model time-varying intensities of events over multiple channels with the CSM kernel that can capture either negative or positive correlations, as well as phase differences between stochastic processes. During the monitoring, the proposed approach measures the level of change in real-time by computing a weighted likelihood of observation with respect to the constructed model and determines whether a target system experiences a change point by conducting a statistical test based on extreme value theory. Our method is validated using three types of datasets: synthetic, accelerometer vibration, and gas regulator data.

Gearbox Fault Prediction of Wind Turbines Based on a Stacking Model and Change-Point Detection

The fault diagnosis and prediction technology of wind turbines are of great significance for increasing the power generation and reducing the downtime of wind turbines. However, most of the current fault detection approaches are realized by setting a single alarm threshold. Considering the complicated working conditions of wind farms, such methods are prone to ignore the fault, send out a false alarm, or leave insufficient troubleshooting time. In this work, we propose a gearbox fault prediction approach of wind turbines based on the supervisory control and data acquisition (SCADA) data. A stacking model composed of Random Forest (RF), Gradient Boosting Decision Tree (GBDT), and Extreme Gradient Boosting (XGBOOST) was constructed as the normal behavior model to describe the normal conditions of the wind turbines. We used the Mahalanobis distance (MD) instead of the residual to measure the deviation of the current state from the normal conditions of the turbines. By inputting the MD series into the proposed change-point detection algorithm, we can obtain the change point at which the fault symptom begins to appear, and thus achieving the fault prediction of the gearbox. The proposed approach is validated on the historical data of 5 wind turbines in a wind farm, which proves its effectiveness to detect the fault in advance.

Edge AI: On-Demand Accelerating Deep Neural Network Inference via Edge Computing

As a key technology of enabling Artificial Intelligence (AI) applications in 5G era, Deep Neural Networks (DNNs) have quickly attracted widespread attention. However, it is challenging to run computation-intensive DNN-based tasks on mobile devices due to the limited computation resources. What's worse, traditional cloud-assisted DNN inference is heavily hindered by the significant wide-area network latency, leading to poor real-time performance as well as low quality of user experience. To address these challenges, in this paper, we propose Edgent, a framework that leverages edge computing for DNN collaborative inference through device-edge synergy. Edgent exploits two design knobs: (1) DNN partitioning that adaptively partitions computation between device and edge for purpose of coordinating the powerful cloud resource and the proximal edge resource for real-time DNN inference; (2) DNN right-sizing that further reduces computing latency via early exiting inference at an appropriate intermediate DNN layer. In addition, considering the potential network fluctuation in real-world deployment, Edgentis properly design to specialize for both static and dynamic network environment. Specifically, in a static environment where the bandwidth changes slowly, Edgent derives the best configurations with the assist of regression-based prediction models, while in a dynamic environment where the bandwidth varies dramatically, Edgent generates the best execution plan through the online change point detection algorithm that maps the current bandwidth state to the optimal configuration. We implement Edgent prototype based on the Raspberry Pi and the desktop PC and the extensive experimental evaluations demonstrate Edgent's effectiveness in enabling on-demand low-latency edge intelligence.