Online Change-point Detection Research Articles

A Data-Driven Framework for FDI Attack Detection and Mitigation in DC Microgrids

This paper proposes a Data-Driven (DD) framework for the real-time monitoring, detection, and mitigation of False Data Injection (FDI) attacks in DC Microgrids (DCMGs). A supervised algorithm is adopted in this framework to continuously estimate the output voltage and current for all Distributed Generators (DGs) with acceptable accuracy. Accordingly, among the various evaluated supervised DD algorithms, Adaptive Neuro-Fuzzy Inference Systems (ANFISs) are utilized because of their low computational burden, efficiency in operation, and simplicity in design and implementation in a distributed control system. The proposed framework is based on the residual analysis of the generated error signal between the estimated and actual sensed signals. The proposed framework detects and mitigates the cyber-attack depending on trends in generated error signals. Moreover, by applying Online Change Point Detection (OCPD), the need for a static user-defined threshold for the residual analysis of the generated error signal is dispelled. Finally, the proposed method is validated in a MATLAB/Simulink testbed, considering the resilience, effectiveness, accuracy, and robustness of multiple case study scenarios.

Quantitative Evaluation of Aerosol Generation During In-Office Flexible Laryngoscopy

Despite growing scientific knowledge and research, it is still unknown if office flexible laryngoscopy (FL) is aerosol generating and thereby potentially increases the risk of SARS-CoV-2 transmission. The limited literature that exists is conflicting, precluding formal conclusions. To determine whether FL is aerosol generating. This prospective cohort study included 134 patients seen in the otolaryngology clinic at a single tertiary care academic institution between February and May 2021. Two optical particle sizer instruments were used, quantifying particles ranging from 0.02 μm to 5 μm. Measurements were taken every 30 seconds, with sample periods of 15 seconds throughout the patient encounter. Instruments were located 12 inches from the patient's nares. Timing of events was recorded, including the start and end of physical examination, topical spray administration, start and end of laryngoscopy, and other potential aerosol-generating events (eg, coughing, sneezing). Data analysis was performed from February to May 2021. Office examination and office FL. Bayesian online change point detection (OCPD) algorithm was used to detect significant change points (CPs) in this time-series data. The primary outcome was significant CP after FL compared with baseline physiologic variations, such as breathing and phonation. Data were collected from 134 patients between February and May 2021. Ninety-one encounters involved FL. Of this group, 51 patients (56%) wore no mask over their mouth during FL. There was no statistically significant CP in either visits involving FL or visits where FL was not performed. Use of nasal spray did not result in CP in aerosol levels. Overall, neither the number of people present in the examination room, masks over patients' mouth, the duration of the visit, nor the duration of FL were associated with mean aerosol counts, regardless of the exposure. For larger aerosol sizes (≥1 μm), however, rooms with higher air exchange rates had significantly higher reductions in mean aerosol counts for visits involving FL. The findings of this cohort study support that FL, including topical spray administration, is not a significant aerosol-generating procedure. The Bayesian OCPD model has a promising application for future aerosol studies in otolaryngology.

Online change-point detection with kernels

Change-points in time series data are usually defined as the time instants at which changes in their properties occur. Detecting change-points is critical in a number of applications as diverse as detecting credit card and insurance frauds, or intrusions into networks. Recently the authors introduced an online kernel-based change-point detection method built upon direct estimation of the density ratio on consecutive time intervals. This paper further investigates this algorithm, making improvements and analyzing its behavior in the mean and mean square sense, in the absence and presence of a change point. These theoretical analyses are validated with Monte Carlo simulations. The detection performance of the algorithm is illustrated through experiments on real-world data and compared to state of the art methodologies.

Sequential changepoint detection in neural networks with checkpoints

We introduce a framework for online changepoint detection and simultaneous model learning which is applicable to highly parametrized models, such as deep neural networks. It is based on detecting changepoints across time by sequentially performing generalized likelihood ratio tests that require only evaluations of simple prediction score functions. This procedure makes use of checkpoints, consisting of early versions of the actual model parameters, that allow to detect distributional changes by performing predictions on future data. We define an algorithm that bounds the Type I error in the sequential testing procedure. We demonstrate the efficiency of our method in challenging continual learning applications with unknown task changepoints, and show improved performance compared to online Bayesian changepoint detection.

High-Dimensional, Multiscale Online Changepoint Detection

AbstractWe introduce a new method for high-dimensional, online changepoint detection in settings where a p-variate Gaussian data stream may undergo a change in mean. The procedure works by performing likelihood ratio tests against simple alternatives of different scales in each coordinate, and then aggregating test statistics across scales and coordinates. The algorithm is online in the sense that both its storage requirements and worst-case computational complexity per new observation are independent of the number of previous observations; in practice, it may even be significantly faster than this. We prove that the patience, or average run length under the null, of our procedure is at least at the desired nominal level, and provide guarantees on its response delay under the alternative that depend on the sparsity of the vector of mean change. Simulations confirm the practical effectiveness of our proposal, which is implemented in the R package ocd, and we also demonstrate its utility on a seismology data set.

Bayesian Online Change Point Detection in Finance

Abstract It is quite common that the structure of a time series changes abruptly. Identifying these change points and describing the model structure in the segments between these change points is an important task in financial time series analysis. Change point detection is the identification of abrupt changes in the generative parameters of sequential data. In application areas such as finance, online rather than offline detection of change points in time series is mostly required, due to their use in predictive tasks, possibly embedded in automatic trading systems. However, the complex structure of the data generation processes makes this a challenging endeavor. This paper is concerned with online change point detection in financial time series using the Bayesian setting. To this end, the Bayesian posterior probability of change at a specific time is proposed and some procedures are presented for selecting the priors and estimation of parameters. Applications in simulated financial time series are given. Finally, conclusions are proposed.

Online Change Point Detection for Weighted and Directed Random Dot Product Graphs

Given a sequence of random (directed and weighted) graphs, we address the problem of online monitoring and detection of changes in the underlying data distribution. Our idea is to endow sequential change-point detection (CPD) techniques with a graph representation learning substrate based on the versatile Random Dot Product Graph (RDPG) model. We consider efficient, online updates of a judicious monitoring function, which quantifies the discrepancy between the streaming graph observations and the nominal RDPG. This reference distribution is inferred via spectral embeddings of the first few graphs in the sequence. We characterize the distribution of this running statistic to select thresholds that guarantee error-rate control, and under simplifying approximations we offer insights on the algorithm’s detection resolution and delay. The end result is a lightweight online CPD algorithm, that is also explainable by virtue of the well-appreciated interpretability of RDPG embeddings. This is in stark contrast with most existing graph CPD approaches, which either rely on extensive computation, or they store and process the entire observed time series. An apparent limitation of the RDPG model is its suitability for undirected and unweighted graphs only, a gap we aim to close here to broaden the scope of the CPD framework. Unlike previous proposals, our non-parametric RDPG model for weighted graphs does not require a priori specification of the weights’ distribution to perform inference and estimation. This network modeling contribution is of independent interest beyond CPD. We offer an open-source implementation of the novel online CPD algorithm for weighted and direct graphs, whose effectiveness and efficiency are demonstrated via (reproducible) synthetic and real network data experiments.

Slow Momentum with Fast Reversion: A Trading Strategy Using Deep Learning and Changepoint Detection

Momentum strategies are an important part of alternative investments and are at the heart of the work of commodity trading advisors. These strategies have, however, been found to have difficulties adjusting to rapid changes in market conditions, such as during the 2020 market crash. In particular, immediately after momentum turning points, when a trend reverses from an uptrend (downtrend) to a downtrend (uptrend), time-series momentum strategies are prone to making bad bets. To improve the responsiveness to regime change, the authors introduce a novel approach, in which they insert an online changepoint detection (CPD) module into a deep momentum network pipeline, which uses a long short-term memory deep-learning architecture to simultaneously learn both trend estimation and position sizing. Furthermore, their model is able to optimize the way in which it balances (1) a slow momentum strategy that exploits persisting trends but does not overreact to localized price moves and (2) a fast mean-reversion strategy regime by quickly flipping its position and then swapping back again to exploit localized price moves. The CPD module outputs a changepoint location and severity score, allowing the model to learn to respond to varying degrees of disequilibrium, or smaller and more localized changepoints, in a data-driven manner. The authors back test their model over the period 1995–2020, and the addition of the CPD module leads to a 33% improvement in the Sharpe ratio. The module is especially beneficial in periods of significant nonstationarity; in particular, over the most recent years tested (2015–2020), the performance boost is approximately 66%. This is especially interesting because traditional momentum strategies underperformed in this period.

Unsupervised Fault Detection With a Decision Fusion Method Based on Bayesian in the Pumping Unit

Since a large amount of data can be obtained in the oil production process nowadays and the operation environment is increasingly complicated, it is necessary to research unsupervised and robust fault detection methods for improving safety. In this paper, an online Bayesian-based technique with a novel decision fusion algorithm is proposed for unsupervised fault detection in the pumping unit. First, a new strategy to detect the working condition of the pumping unit by dynamometer card as well as five process measured variables is proposed. To deal with high-dimension data and outliers in dynamometer card, a robust Douglas-Peucker algorithm is developed for obtaining compressed data. A chord ratio index evaluating deviation degree of observations is defined, which can be used for removing outliers during approximation. Two norms are introduced for choosing the threshold in the proposed Douglas-Peucker algorithm. Moreover, a Bayesian-based online change point detection model is attempted for detecting univariate faults in the pumping unit. A decision fusion method derived from Bayesian probability formula is proposed for fusing univariate fault detection results. At last, the power of the proposed method is evaluated by numerical simulations and a real oil production process.

Data-driven crack behavior anomaly identification method for concrete dams in long-term service using offline and online change point detection

Data-driven crack behavior anomaly identification method for concrete dams in long-term service using offline and online change point detection

TCP-NeuRoc: Neural Adaptive TCP Congestion Control With Online Changepoint Detection

Congestion control is a fundamental mechanism for TCP protocol, which has been extensively studied in the past three decades. However, our experimental evaluations show that the state-of-art congestion control algorithms such as Cubic and BBR are far from optimal: they have unresolved issues such as insufficient usage of available bandwidth, inadaptable to dynamic bandwidth variants, and compromising on one or more performance dimensions. To address these challenges, we propose a novel congestion control mechanism called NeuRoc that coordinately uses online changepoint detection and deep reinforcement learning (DRL) technique to generate the optimal congestion control policy, which allows TCP operating at Kleinrock ’s optimal operation point to achieve fully bandwidth usage and low latency. To address the practical issues of deploying the deep learning based congestion control mechanism, we propose a cold-started training and deployment framework to reduce the cost of bootstrap. We implement NeuRoc on an emulation platform which connects to the Linux network protocol stack through virtual network interfaces. Extensive experiments show that NeuRoc achieves the best throughput-latency tradeoff compared with the state-of-the-arts in a variety of scenarios.

The role of parameters in Bayesian Online Changepoint Detection: detecting early warning of mount Merapi eruptions

Indonesia is a country that is surrounded by active volcanoes, which may erupt at any time; therefore, an online early warning system of volcanic eruption is crucial. In this paper, an online early warning system is constructed based on the changepoints detection on earthquake magnitude time series. This online early warning system is built using a Bayesian Online Changepoint Detection (BOCPD) method. One of the method's advantages is that one can customize the parameters (initial hyper-parameters and hazard-rate parameter) of BOCPD to follow a chosen constraint. These parameters determine the time and number of changepoints. An algorithm, called Appropriate Parameters of Bayesian Online Changepoint Detection for Early Warning (APBOCPD-EW), is proposed to get the parameters that lead the detection to the early warning points before eruption. We apply the algorithm for online early warning of mount Merapi eruptions. The results show that the proposed method produces parameters that give good estimation time for early warnings of mount Merapi's eruptions.

The Bethe Hessian and Information Theoretic Approaches for Online Change-Point Detection in Network Data

The Bethe Hessian and Information Theoretic Approaches for Online Change-Point Detection in Network Data

Bayesian Online Change Point Detection for Baseline Shifts

In time series data analysis, detecting change points on a real-time basis (online) is of great interest in many areas, such as finance, environmental monitoring, and medicine. One promising means to achieve this is the Bayesian online change point detection (BOCPD) algorithm, which has been successfully adopted in particular cases in which the time series of interest has a fixed baseline. However, we have found that the algorithm struggles when the baseline irreversibly shifts from its initial state. This is because with the original BOCPD algorithm, the sensitivity with which a change point can be detected is degraded if the data points are fluctuating at locations relatively far from the original baseline. In this paper, we not only extend the original BOCPD algorithm to be applicable to a time series whose baseline is constantly shifting toward unknown values but also visualize why the proposed extension works. To demonstrate the efficacy of the proposed algorithm compared to the original one, we examine these algorithms on two real-world data sets and six synthetic data sets.

Two-stage physics-based Wiener process models for online RUL prediction in field vibration data

Due to most failure mechanisms, such as fatigue crack growth and fatigue spall, the degradation process of rotating machinery commonly exhibits two-stage features in engineering practice. Other minor factors are also the key issues affecting the health evolution process, including the component structure, assembly accuracy, and working environment. Ignoring such a mechanism may lead to imprecise in degradation modeling, life prognostic, and ultimately lead to safety risk. Besides, achieving high accuracy of prognostic emphasizes the influence of random effect in the degradation process. The contribution of this study lies in addressing this issue by proposing two-stage physics-based Wiener process models integrating: (a) fatigue crack mechanism and crack growth law, and (b) other minor factors. A general prognostic framework is formulated by jointly employing the online change point detection, parameter estimation, and remaining useful life (RUL) prediction, which has good statistic inference and applicability in two general nonlinear systems, i.e., power-law and exponential-law. A joint implement of offline two-step parameter estimation method and the online Bayesian update method is executed, making full advantage of historical and in-service data, based on which the RUL prediction transcends into an imperative PHM module. A practical case study on the vibration dataset of wheel treads demonstrates the practically implement ability of the proposed method in achieving high accuracy of RUL prediction.

Disentangling conditional effects of multiple regime shifts on Atlantic cod productivity.

Regime shifts are increasingly prevalent in the ecological literature. However, definitions vary and detection methods are still developing. Here, we employ a novel statistical algorithm based on the Bayesian online change-point detection framework to simultaneously identify shifts in the mean and (or) variance of time series data. We detected multiple regime shifts in long-term (59–154 years) patterns of coastal Norwegian Atlantic cod (>70% decline) and putative drivers of cod productivity: North Atlantic Oscillation (NAO); sea-surface temperature; zooplankton abundance; fishing mortality (F). The consequences of an environmental or climate-related regime shift on cod productivity are accentuated when regime shifts coincide, fishing mortality is high, and populations are small. The analyses suggest that increasing F increasingly sensitized cod in the mid 1970s and late 1990s to regime shifts in NAO, zooplankton abundance, and water temperature. Our work underscores the necessity of accounting for human-induced mortality in regime shift analyses of marine ecosystems.

Two multivariate online change detection models

ABSTRACT Online change point detection methods monitor changes in the distribution of a data stream. This article discusses two non-parametric online change detection methods based on the energy statistics and Mahalanobis depth. To apply the energy statistic, we use sliding-window algorithm with efficient training and updating procedures. For Mahalanobis depth, we propose an algorithm to train the threshold with desired protective ability against false alarms and discuss factors that have an influence on the threshold. Numerical studies evaluate and compare the performance of the proposed models with three existing methods to detect changes in the mean and variability of a data stream. The methods are applied to detecting changes in the flowing volume of the Mississippi River.

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.

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.

NEWMA: A New Method for Scalable Model-Free Online Change-Point Detection

We consider the problem of detecting abrupt changes in the distribution of a multi-dimensional time series, with limited computing power and memory. In this paper, we propose a new, simple method for model-free online change-point detection that relies only on fast and light recursive statistics, inspired by the classical Exponential Weighted Moving Average algorithm (EWMA). The proposed idea is to compute two EWMA statistics on the stream of data with different forgetting factors, and to compare them. By doing so, we show that we implicitly compare recent samples with older ones, without the need to explicitly store them. Additionally, we leverage Random Features (RFs) to efficiently use the Maximum Mean Discrepancy as a distance between distributions, furthermore exploiting recent optical hardware to compute high-dimensional RFs in near constant time. We show that our method is significantly faster than usual non-parametric methods for a given accuracy.