Online Change Detection Research Articles

Bayesian Online Change Detection of Multimode Processes with Application to Wind Turbines

Bayesian Online Change Detection of Multimode Processes with Application to Wind Turbines

Detecting Abrupt Change in Channel Covariance Matrix for MIMO Communication

The acquisition of the channel covariance matrix is of paramount importance to many strategies in multiple-input-multiple-output (MIMO) communications, such as the minimum mean-square error (MMSE) channel estimation. Therefore, plenty of efficient channel covariance matrix estimation schemes have been proposed in the literature. However, an abrupt change in the channel covariance matrix may happen occasionally in practice due to the change in the scattering environment and the user location. Our paper aims to adopt the classic change detection theory to detect the change in the channel covariance matrix as accurately and quickly as possible such that the new covariance matrix can be re-estimated in time. Specifically, this paper first considers the technique of on-line change detection (also known as quickest/sequential change detection), where we need to detect whether a change in the channel covariance matrix occurs at each channel coherence time interval. Next, because the complexity of detecting the change in a high-dimension covariance matrix at each coherence time interval is too high, we devise a low-complexity off-line strategy in massive MIMO systems, where change detection is merely performed at the last channel coherence time interval of a given time period. Numerical results show that our proposed on-line and off-line schemes can detect the channel covariance change with a small delay and a low false alarm rate. Therefore, our paper theoretically and numerically verifies the feasibility of detecting the channel covariance change accurately and quickly in practice.

Spectral CUSUM for Online Network Structure Change Detection

Detecting abrupt changes in the community structure of a network from noisy observations is a fundamental problem in statistics and machine learning. This paper presents an online change detection algorithm called Spectral-CUSUM to detect unknown network structure changes through a generalized likelihood ratio statistic. We characterize the average run length (ARL) and the expected detection delay (EDD) of the Spectral-CUSUM procedure and prove its asymptotic optimality. Finally, we demonstrate the good performance of the Spectral-CUSUM procedure and compare it with several baseline methods using simulations and real data examples on seismic event detection using sensor network data.

Monitoring Land Cover Change by Leveraging a Dynamic Service-Oriented Computing Model

Land cover change (LCC) is increasingly affecting global climate change, energy cycle, carbon cycle, and water cycle, with far-reaching consequences to human well-being. Web service-based online change detection applications have bloomed over the past decade for monitoring land cover change. Currently, massive processing services and data services have been published and used over the internet. However, few studies consider both service integration and resource sharing in land cover domain, making end-users rarely able to acquire the LCC information timely. The behavior interaction between services is also growing more complex due to the increasing use of web service composition technology, making it challenging for static web services to provide collaboration and matching between diverse web services. To address the above challenges, a Dynamic Service Computing Model (DSCM) was proposed for monitoring LCC. Three dynamic computation strategies were proposed according to different users’ requirements of change detection. WMS-LCC was first developed by extending the existing WMS for ready-use LCC data access. Spatial relation-based LCC data integration was then proposed for extracting LCC information based on multi-temporal land cover data. Processing service encapsulation and service composition methods were also developed for chaining various land cover services to a complex service chain. Finally, a prototype system was implemented to evaluate the validity and feasibility of the proposed DSCM. Two walk-through examples were performed with GlobeLand30 datasets and muti-temporal Landsat imagery, respectively. The experimental results indicate that the proposed DSCM approach was more effective and applicable to a wider range of issues in land cover change detection.

Domain Constraints-Driven Automatic Service Composition for Online Land Cover Geoprocessing

With the rapid development of web service technology, automatic land cover web service composition has become one of the key challenges in solving complex geoprocessing tasks of land cover. Service composition requires the creation of service chains based on semantic information about the services and all the constraints that should be respected. Artificial intelligence (AI) planning algorithms have recently significantly progressed in solving web service composition problems. However, the current approaches lack effective constraints to guarantee the accuracy of automatic land cover service composition. To address this challenge, the paper proposes a domain constraints-driven automatic service composition approach for online land cover geoprocessing. First, a land cover service ontology was built to semantically describe land cover tasks, data, and services, which assist in constructing domain constraints. Then, a constraint-aware GraphPlan algorithm was proposed, which constructs a service planning graph and searches services based on the domain constraints for generating optimal web service composition solutions. In this paper, the above method was integrated into a web prototype system and a case study for the online change detection automatic geoprocessing was implemented to test the accuracy of the method. The experimental results show that with this method, a land cover service chain can generate automatically by user desire objective and domain constraints, and the service chain execution result is more accurate.

A Bayesian Partially Observable Online Change Detection Approach with Thompson Sampling

This article proposes a Bayesian learning framework for online change detection of high-dimensional data streams where only a subset of variables can be observed at each time point due to limited sensing capacities. On the one hand, we need to build a change detection scheme based on partial observations. On the other, the scheme should be able to adaptively and actively select the most critical sensing variables to observe to maximize the detection power. To address these two points, in this article, first, a novel Bayesian Spike-Slab Composite Decomposition (BSSCD) is proposed to decompose the high-dimensional signals onto normal and abnormal bases, where the projection coefficients are efficiently estimated via variational Bayesian inference. Built upon it, the posterior Bayes factor is constructed as the detection statistic. Second, by further formulating the detection statistic as the reward function of combinatorial multi-armed bandit (CMAB), a Thompson sampling strategy is proposed for selecting the potential changed variables with the balance of exploration and exploitation. The efficacy and applicability of our method are demonstrated in practice with numerical studies and a real case study.

Online Forest Disturbance Detection at the Sub-Annual Scale Using Spatial Context From Sparse Landsat Time Series

Mapping forest disturbances using dense time series can timely identify disturbances at the subannual scale. However, these change detection methods using dense time series may be infeasible when not enough temporal observations are available. In this article, an online change detection algorithm that identifies forest disturbances at a subannual scale using spatial context from the sparse Landsat time series was proposed. First, the spatial normalized index that removed forest seasonality was prepared for establishing a simplified model instead of the harmonic model, thereby reducing the requirements for a high temporal frequency of clear observations for model initialization. Second, by using the spatial errors model to establish the simplified model, the normally distributed residual time series that removed the spatial autocorrelation were obtained. Third, the spatial statistic <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$t$ </tex-math></inline-formula> time series transformed from residual time series within a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$3\times3$ </tex-math></inline-formula> spatial window were subsequently subjected to the exponentially weighted moving average <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$t$ </tex-math></inline-formula> chart (EWMA-t), which is a statistical process control chart for a short cycle corresponding to sparse Landsat time series. Fourth, disturbed pixels were labeled if the chart values persistently deviated from the control limits of the chart. The proposed algorithm was applied to a subtropical forest with low Landsat data availability and yielded an overall accuracy of 86% in the spatial domain and temporal accuracy of 93.7%, achieving accurate and timely identification of forest disturbances. The proposed method called the EWMA-t change detection (EWMATCD) algorithm provides an alternative for disturbance detection at the subannual scale in regions with low data availability.

Nonparametric and Online Change Detection in Multivariate Datastreams Using QuantTree

We address the problem of online change detection in multivariate datastreams, and we introduce QuantTree Exponentially Weighted Moving Average (QT-EWMA), a nonparametric change-detection algorithm that can control the expected time before a false alarm, yielding a desired Average Run Length (ARL$_0$). Controlling false alarms is crucial in many applications and is rarely guaranteed by online change-detection algorithms that can monitor multivariate datastreams without knowing the data distribution. Like many change-detection algorithms, QT-EWMA builds a model of the data distribution, in our case a QuantTree histogram, from a stationary training set. To monitor datastreams even when the training set is extremely small, we propose QT-EWMA-update, which incrementally updates the QuantTree histogram during monitoring, always keeping the ARL$_0$ under control. Our experiments, performed on synthetic and real-world datastreams, demonstrate that QT-EWMA and QT-EWMA-update control the ARL$_0$ and the false alarm rate better than state-of-the-art methods operating in similar conditions, achieving lower or comparable detection delays.

CUSUM multi-chart based on nonparametric likelihood approach for detecting unknown abrupt changes and its application for network data

This article deals with the online monitoring of changes when the distributions (including the form of distributions) of pre-change and post-change are unknown. We not only construct a nonparametric CUSUM multi-chart based on the nonparametric likelihood function to deal with the online change detection problem, but also give a strategy for choosing the appropriate value if we choose only one truncated value of u. And under the measurement called nonparametric Control chart Performance Index (CPI), we show an asymptotic optimal design for the allocation of the reference post-change Cumulative Distribution Functions (C.D.F.s). Finally, we use simulated data to present the related results. Moreover, we apply the method in this paper to network data, which is very common at present.

Online common change-point detection in a set of nonstationary categorical time series

Categorical sequences are widely used in various domains to describe the evolutionary state of the process under study. This article addresses the problem of behavioral change detection for multiple categorical time series. Relying on the sequential likelihood ratio test, an online change detection method is proposed based on the joint modeling of all the categorical sequences. To model the joint probability density, a nonhomogeneous Markov model is used. It allows modeling the transition dynamics over time and considering their dependence on some exogenous factors that may influence the behavior changes. An adaptive threshold is learned using Monte Carlo simulations to detect different changes and reduce false alarms. The performance of the proposed method is evaluated using two real-world and four synthetic datasets. It is compared with two state-of-the-art change detection methods, namely logistic regression and homogeneous Markov model. The experimentation using synthetic datasets highlights the proposed method’s effectiveness in terms of both the detection precision and the detection delay. The real-world data are issued from a water network and school-to-work transition. The analysis of the model estimated parameters allows us to characterize the detected changes in a real-world context.

Online droplet anomaly detection from streaming videos in inkjet printing

Inkjet printing (IJP) has demonstrated its capabilities to produce high-quality and high-resolution parts, such as sensors, bio-chips, etc., with outstanding functionality. However, the quality of the printed parts can be endangered by the abnormal droplet jetting behaviors, which are substantially governed by the process dynamics and ambient conditions. Additionally, the droplet behaviors extensively define the final drop deposition quality in IJP. Timely capturing and identifying abrupt anomalies that the ejected droplets may suffer from are primordial for the quality assurance of the printed parts. Machine vision systems are able to record the droplet videos during IJP. Nevertheless, it is challenging to timely detect the anomalies from the collected droplet videos. The objective of this work is to build an analytical framework that allows for online droplet anomaly detection from droplet videos. There are several challenges that have not been addressed before: (1) the features of the droplet videos need to be extracted in an online fashion; and (2) there is no well-defined baseline to support the droplet anomalies detection, due to the complex process dynamics. Here, we propose a novel online framework to efficiently detect the anomalies from process streaming videos. In particular, we extend the multivariate Bayesian online change detection (BOCD) framework to high-dimensional data (i.e., tensor data of droplet videos) by leveraging online tensor factorization (OTF). OTF decomposes the streaming data into non-temporal and temporal low-dimensional factorization matrices. The non-temporal factorization matrices are deployed to extract the frame-specific temporal factorization matrix within a user-defined sliding window. Subsequently, the temporal factorization matrix for each frame is monitored with BOCD, which accurately detects anomalies in the streaming data. The proposed framework is demonstrated by detecting droplet anomalies from streaming data in IJP, showing excellent accuracy and efficiency.

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 Forecasting Approach to Online Change Detection in Land Cover Time Series

We present a method for online detection of land cover change based on remotely sensed time series. Change is detected by monitoring deviations between observations and forecasts made using the time series historical data and similar time series in the geographical region. This method and several others were applied to MODIS 8-day surface reflectance data for problems of detecting settlement expansion in Limpopo Province, South Africa, and detecting deforestation in New South Wales, Australia. The proposed method had significantly shorter median detection delay (DD) for equivalent rates of false alarms compared with the other evaluated methods. We obtained a median DD of seven samples for settlement detection and 14 samples for deforestation detection corresponding to 56 days and 112 days, respectively. This is compared with a median DD of 224 and 544 days for the best other methods evaluated. We suggest that the proposed method is an excellent candidate for land cover change detection where rapid detection is essential.

Adaptive long-term traffic state estimation with evolving spiking neural networks

Due to the nature of traffic itself, most traffic forecasting models reported in literature aim at producing short-term predictions, yet their performance degrades when the prediction horizon is increased. The scarce long-term estimation strategies currently found in the literature are commonly based on the detection and assignment to patterns, but their performance decays when unexpected events provoke non predictable changes, or if the allocation to a traffic pattern is inaccurate. This work introduces a method to obtain long-term pattern forecasts and adapt them to real-time circumstances. To this end, a long-term estimation scheme based on the automated discovery of patterns is proposed and integrated with an on-line change detection and adaptation mechanism. The framework takes advantage of the architecture of evolving Spiking Neural Networks (eSNN) to perform adaptations without retraining the model, allowing the whole system to work autonomously in an on-line fashion. Its performance is assessed over a real scenario with 5 min data of a 6-month span of traffic in the center of Madrid, Spain. Significant accuracy gains are obtained when applying the proposed on-line adaptation mechanism on days with special, non-predictable events that degrade the quality of their long-term traffic forecasts.

An empirical likelihood-based CUSUM for on-line model change detection

In change detection problem, the distribution of a series of observations can change at some unknown instant. The aim of on-line change detection rule is to detect this change, as rapidly as possible, while ensuring a low rate of false alarm. The most popular rule to treat this problem is the Page’s CUSUM rule. The use of this rule supposes that the two distributions, before and after the change, are known, which is often restrictive in practice. In this article, a nonparametric rule is proposed. Only two learning samples, characterizing the in-control and the out-of-control functioning modes of the system, are needed to implement the rule. The new detection approach is based on the use of a well-known nonparametric method, Empirical Likelihood. Some numerical studies show the relevance of our approach, especially when the size of the learning samples are quite small.

Sensor-Based Change Detection for Timely Solicitation of User Engagement

The accurate detection of changes has the potential to form a fundamental component of systems which autonomously solicit user interaction based on transitions within an input stream, for example, electrocardiogram data or accelerometry obtained from a mobile device. This solicited interaction may be utilized for diverse scenarios such as responding to changes in a patient's vital signs within a medical domain or requesting user activity labels for generating real-world labelled datasets. Within this paper, we extend our previous work on the Multivariate Online Change detection Algorithm subsequently exploring the utility of incorporating the Benjamini Hochberg method of correcting for multiple comparisons. Furthermore, we evaluate our approach against similarly light-weight Multivariate Exponentially Weighted Moving Average and Cumulative Sum based techniques. Results are presented based on manually labelled change points in accelerometry data captured using 10 participants. Each participant performed nine distinct activities for a total period of 35 minutes. The results subsequently demonstrate the practical potential of our approach from both accuracy and computational perspectives.

A service relation model for web-based land cover change detection

Change detection with remotely sensed imagery is a critical step in land cover monitoring and updating. Although a variety of algorithms or models have been developed, none of them can be universal for all cases. The selection of appropriate algorithms and construction of processing workflows depend largely on the expertise of experts about the “algorithm-data” relations among change detection algorithms and the imagery data used. This paper presents a service relation model for land cover change detection by integrating the experts’ knowledge about the “algorithm-data” relations into the web-based geo-processing. The “algorithm-data” relations are mapped into a set of web service relations with the analysis of functional and non-functional service semantics. These service relations are further classified into three different levels, i.e., interface, behavior and execution levels. A service relation model is then established using the Object and Relation Diagram (ORD) approach to represent the multi-granularity services and their relations for change detection. A set of semantic matching rules are built and used for deriving on-demand change detection service chains from the service relation model. A web-based prototype system is developed in .NET development environment, which encapsulates nine change detection and pre-processing algorithms and represents their service relations asan ORD. Three test areas from Shandong and Hebei provinces, China with different imagery conditions are selected for online change detection experiments, and the results indicate that on-demand service chains can be generated according to different users’ demands.

Change detection for uncertain autoregressive dynamic models through nonparametric estimation

A new statistical approach for on-line change detection in uncertain dynamic system is proposed. In change detection problem, the distribution of a sequence of observations can change at some unknown instant. The goal is to detect this change, for example a parameter change, as quickly as possible with a minimal risk of false detection. In this paper, the observations come from an uncertain system modeled by an autoregressive model containing an unknown functional component. The popular Page’s CUSUM rule is not applicable anymore since it requires the full knowledge of the model. A new detection CUSUM-like scheme is proposed, which is based on the nonparametric estimation of the unknown component from a learning sample. Moreover, the estimation procedure can be updated on-line which ensures a better detection, especially at the beginning of the monitoring procedure. Simulation trials were performed on a model describing a water treatment process and show the interest of this new procedure with respect to the classic CUSUM rule.

Dynamic detection of window starting positions and its implementation within an activity recognition framework

Activity recognition is an intrinsic component of many pervasive computing and ambient intelligent solutions. This has been facilitated by an explosion of technological developments in the area of wireless sensor network, wearable and mobile computing. Yet, delivering robust activity recognition, which could be deployed at scale in a real world environment, still remains an active research challenge. Much of the existing literature to date has focused on applying machine learning techniques to pre-segmented data collected in controlled laboratory environments. Whilst this approach can provide valuable ground truth information from which to build recognition models, these techniques often do not function well when implemented in near real time applications. This paper presents the application of a multivariate online change detection algorithm to dynamically detect the starting position of windows for the purposes of activity recognition.

Incorporating change detection in the monitoring phase of adaptive query processing

Recent Big Data research typically emphasizes on the need to address the challenges stemming from the volume, velocity, variety and veracity aspects. However, another cross-cutting property of Big Data is volatility. In database technology, volatility is addressed with the help of adaptive query processing (AQP), which has become the dominant paradigm for executing queries in dynamic and/or streaming environments. As the characteristics of the runtime environment may vary significantly along time, AQP techniques employ a three-phase adaptivity loop to process the input queries, comprising feedback collection, analysis and re-optimization. In the monitoring phase, the standard approach is to collect feedback in a fixed-size sliding window. However, several problems arise when the techniques adopt a fixed-size sliding window for maintaining runtime collected feedback. In this work, we tackle this limitation and we propose a novel monitoring phase, which assesses the collected feedback rendering an AQP technique capable of taking more informed decisions during the subsequent phases. The proposed approach is non-intrusive to the state-of-the-art adaptivity loop and can adopt any state-of-the-art online change detection algorithm through its plug-and-play abstraction. Another contribution of this work is a novel algorithm for detecting changes in a filter’s drop probability, called β-CUSUM. The potential of the novel monitoring phase and of β-CUSUM is experimentally evaluated using both real-world and synthetic datasets.