Change Point Detection Research Articles

Spatio-Temporal Dynamics of Total Suspended Sediments in the Belize Coastal Lagoon

Increased tourism in Belize over the last decade and the growth of the local population have led to coastal development and infrastructure expansion. Land use alteration and anthropogenic activity may change the sediment and nutrient loads in coastal systems, which can negatively affect ecosystems via mechanisms such as reducing photosynthetically active radiation fields, smothering sessile habitats, and stimulating eutrophication events. Accurate monitoring and prediction of water quality parameters such as Total Suspended Sediments (TSS), are essential in order to understand the influence of land-based changes, climate, and human activities on the coastal systems and devise strategies to mitigate negative impacts. This study implements machine learning algorithms such as Random Forests (RF), Extreme Gradient Boosting (XGB), and Deep Neural Networks (DNN) to estimate TSS using Sentinel-2 reflectance data in the Belize Coastal Lagoon (BCL) and validates the results using TSS data collected in situ. DNN performed the best and estimated TSS with a testing R2 of 0.89. Time-series analysis was also performed on the BCL’s TSS trends using Bayesian Changepoint Detection (BCD) methods to flag anomalously high TSS spatio-temporally, which may be caused by dredging events. Having such a framework can ease the near-real-time monitoring of water quality in Belize, help track the TSS dynamics for anomalies, and aid in meeting and maintaining the sustainable goals for Belize.

Predictive modeling for online in-plane shape deviation inspection and compensation of additive manufacturing

PurposeLimited geometric accuracy is one of the major challenges that hinder the wider application of additive manufacturing (AM). This paper aims to predict in-plane shape deviation for online inspection and compensation to prevent error accumulation and improve shape fidelity in AM.Design/methodology/approachA sequence-to-sequence model with an attention mechanism (Seq2Seq+Attention) is proposed and implemented to predict subsequent layers or the occluded toolpath deviations after the multiresolution alignment. A shape compensation plan can be performed for the large deviation predicted.FindingsThe proposed Seq2Seq+Attention model is able to provide consistent prediction accuracy. The compensation plan proposed based on the predicted deviation can significantly improve the printing fidelity for those layers detected with large deviations.Practical implicationsBased on the experiments conducted on the knee joint samples, the proposed method outperforms the other three machine learning methods for both subsequent layer and occluded toolpath deviation prediction.Originality/valueThis work fills a research gap for predicting in-plane deviation not only for subsequent layers but also for occluded paths due to the missing scanning measurements. It is also combined with the multiresolution alignment and change point detection to determine the necessity of a compensation plan with updated G-code.

Research on time series change point detection and influencing factors under machine learning: based on PM2.5 concentration data in Hefei city

Research on time series change point detection and influencing factors under machine learning: based on PM2.5 concentration data in Hefei city

Detecting Global Events with Bayesian Changepoint Detection on Flight Data

As a means of enhancing the intelligence community’s ability to be prepared and prompt in the detection of global events, we present a Bayesian changepoint detection methodology to detect anomalous activity in open-source flight data. We demonstrate a computationally inexpensive methodology to monitor a geographic location for changes in flight activity over time and demonstrate its utility in a case study of the 2022 Russian invasion of Ukraine. Beyond flight data, the methodology described in this paper shows the potential to generate near real-time situational awareness using open-source time series data streams. This paper was previously published and presented in the Donald R. Keith Memorial Capstone Conference at USMA in May of 2023.

Supervised Learning of Physical Activity Features From Functional Accelerometer Data.

We propose a new health informatics framework to analyze physical activity (PA) from accelerometer devices. Accelerometry data enables scientists to extract personal digital features useful for precision health decision making. Existing methods in accelerometry data analysis typically begin with discretizing summary counts by certain fixed cutoffs into activity categories. One well-known limitation is that the chosen cutoffs are often validated under restricted settings, and cannot be generalizable across populations, devices, or studies. We develop a data-driven approach to overcome this bottleneck in PA data analysis, in which we holistically summarize a subject's activity profile using Occupation-Time curves (OTCs), which describe the percentage of time spent at or above a continuum of activity count levels. We develop multi-step adaptive learning algorithms to perform supervised learning via a scalar-on-function model that involves OTC as the functional predictor of interest as well as other scalar covariates. Our learning analytic first incorporates a hybrid approach of fused lasso for clustering and Hidden Markov Model for changepoint detection, then executes refinement procedures to determine activity windows of interest. We evaluate and illustrate the performance of the proposed learning analytic through simulation experiments and real-world data analyses to assess the influence of PA on biological aging. Our findings indicate a different directional relationship between biological age and PA depending on the specific outcome of interest. Our bioinformatics methodology involves the biomedical outcome of interest to detect different critical points, and is thus adaptive to the specific data, study population, and health outcome under investigation.

Interpretation of a change point detection problem via adaptive lasso penalty and its application

Interpretation of a change point detection problem via adaptive lasso penalty and its application

Network representations of attractors for change point detection

A common approach to monitoring the status of physical and biological systems is through the regular measurement of various system parameters. Changes in a system’s underlying dynamics manifest as changes in the behaviour of the observed time series. For example, the transition from healthy cardiac activity to ventricular fibrillation results in erratic dynamics in measured electrocardiogram (ECG) signals. Identifying these transitions—change point detection—can be valuable in preparing responses to mitigate the effects of undesirable system changes. Here, we present a data-driven method of detecting change points using a phase space approach. Delay embedded trajectories are used to construct an ‘attractor network’, a discrete Markov-chain representation of the system’s attractor. Once constructed, the attractor network is used to assess the level of surprise of future observations where unusual movements in phase space are assigned high surprise scores. Persistent high surprise scores indicate deviations from the attractor and are used to infer change points. Using our approach, we find that the attractor network is effective in automatically detecting the onset of ventricular fibrillation (VF) from observed ECG data. We also test the flexibility of our method on artificial data sets and demonstrate its ability to distinguish between normal and surrogate time series.

A class of smooth, possibly data-adaptive nonparametric copula estimators containing the empirical beta copula

A broad class of smooth, possibly data-adaptive nonparametric copula estimators that contains empirical Bernstein copulas introduced by Sancetta and Satchell (and thus the empirical beta copula proposed by Segers, Sibuya and Tsukahara) is studied. Within this class, a subclass of estimators that depend on a scalar parameter determining the amount of marginal smoothing and a functional parameter controlling the shape of the smoothing region is specifically considered. Empirical investigations of the influence of these parameters suggest to focus on two particular data-adaptive smooth copula estimators that were found to be uniformly better than the empirical beta copula in all of the considered Monte Carlo experiments. Finally, with future applications to change-point detection in mind, conditions under which related sequential empirical copula processes converge weakly are provided.

DDoS attack forecasting based on online multiple change points detection and time series analysis

DDoS attack forecasting based on online multiple change points detection and time series analysis

What makes a real change in software performance? An empirical study on analyzing the factors that affect the triagement of performance change points

Performance tests can be automatically done on each commit version with a continuous integration system. After code changes are pushed to the repository, multiple benchmark tests are run to measure system's performance. Change point detection technology analyzes all these outcomes to identify commits that significantly change performance automatically. However, a considerable number of automatically detected change points are tagged as not being actionable (false positive). Typically, validating potential change points by hand is difficult, time-consuming, and thus creates a bottleneck in the testing process.Our work focuses on identifying what factors affect performance change point triagement and triaging whether a newly detected change point is true positive or not automatically. We start by extracting 34 features from four dimensions i.e., Configuration, Time Series, Version, and Context. We utilize a random forest classifier to triage the change points based on the proposed features. The results indicate that most of the proposed features differ significantly between true positive and false positive change points. On average, our model obtains an AUC of 0.881, which is statistically considerably better than two state-of-the-art approaches. We also look into the most important features for triagement and discover that version related features are the most essential factors.

A Bayesian change-point detection approach to the economic evaluation of risky projects: an application to healthcare technology assessment

Abstract We propose a Bayesian hypothesis testing framework that allows for the assessment of evidence collected during a clinical trial about the cost-effectiveness of a healthcare technology. The model exploits a Bayesian updating rule that makes the link between the evidence collected in clinical research and the expected payoffs of adoption to the healthcare system. The framework takes into account the cost of decision errors in the payoff function, allowing the decision maker to compute the cost of taking a decision when evidence is far from the optimal decision triggers. We show, using a real-world cost-effectiveness study based on clinical trial evidence, how rules derived from a sequential adaptive design approach can lead to quicker decisions when compared to the value of information decision framework. Our application shows that a sequential approach has the potential to lead to quicker decisions, higher payoffs, and better health outcomes.

Data-adaptive symmetric CUSUM for sequential change detection

Detecting change points sequentially in a streaming setting, especially when both the mean and the variance of the signal can change, is often a challenging task. A key difficulty in this context often involves setting an appropriate detection threshold, which for many standard change statistics may need to be tuned depending on the prechange and postchange distributions. This presents a challenge in a sequential change detection setting when a signal switches between multiple distributions. Unfortunately, change point detection schemes that use the log-likelihood ratio, such as cumulative sum (CUSUM) and the generalized log-likelihood ratio (GLR), are quick to react to changes but are not symmetric when both the mean and the variance of the signal change. This makes it difficult to set a single threshold to detect multiple change points sequentially in a streaming setting. We propose a modified version of CUSUM that we call data-adaptive symmetric CUSUM (DAS-CUSUM). The DAS-CUSUM procedure is symmetric for changes between distributions, making it suitable to set a single threshold to detect multiple change points sequentially in a streaming setting. We provide results that relate the expected detection delay and average run length for our proposed procedure when both prechange and postchange distributions are normally distributed. Experiments on simulated and real-world data show the utility of DAS-CUSUM.

Leak State Detection and Size Identification for Fluid Pipelines with a Novel Acoustic Emission Intensity Index and Random Forest.

In this paper, an approach to perform leak state detection and size identification for industrial fluid pipelines with an acoustic emission (AE) activity intensity index curve (AIIC), using b-value and a random forest (RF), is proposed. Initially, the b-value was calculated from pre-processed AE data, which was then utilized to construct AIICs. The AIIC presents a robust description of AE intensity, especially for detecting the leaking state, even with the complication of the multi-source problem of AE events (AEEs), in which there are other sources, rather than just leaking, contributing to the AE activity. In addition, it shows the capability to not just discriminate between normal and leaking states, but also to distinguish different leak sizes. To calculate the probability of a state change from normal condition to leakage, a changepoint detection method, using a Bayesian ensemble, was utilized. After the leak is detected, size identification is performed by feeding the AIIC to the RF. The experimental results were compared with two cutting-edge methods under different scenarios with various pressure levels and leak sizes, and the proposed method outperformed both the earlier algorithms in terms of accuracy.

Increased troponin I blood level and in-hospital outcome across successive COVID-19 waves

Abstract Introduction Troponin I blood level (TnI) is often elevated in patients hospitalized due to COVID-19, and high TnI is associated with adverse outcomes, in this scenario. However, the case rate of high TnI in hospitalized patients and respective case fatality rates (CFR) across different COVID-19 waves still demand investigation, as emerging SARS-CoV-2 variants may affect the cardiac tissue diversely. This study investigated the case rate of high TnI and respective in-hospital outcome in patients hospitalized due to COVID-19 during successive pandemic waves. Methods In a retrospective observational cohort study, patients hospitalized due to COVID-19 from March 1st 2020 to July 31st 2022 in a reference hospital were analyzed. Inclusion criteria were: positive SARS-CoV-2 RT-PCR, hospitalization in a COVID-19 cohort ward or ICU, and at least one TnI assessed during hospitalization. Patients presenting acute coronary syndromes were excluded. TnI was assessed using electrochemiluminescence technique, and the highest value during hospital stay considered for analysis. TnI was positive (+) if above normality reference level and negative (-), otherwise. Outcome was in-hospital survival. Age on admission, total length-of-stay (TLOS), TnI(+) case rate, and CFR (the rate of non-survivors to COVID-19 admissions) were analyzed. COVID-19 waves were analyzed on 7-day moving average positive cases series (1), and respective onset and offset assessed by change point detection algorithm and visual inspection. In each wave period, dominants circulating variants from surveillance data (2,3) were matched. Discrete variables were described as percentage and analyzed by Chi-squared test. Continuous variables were described as mean ± SD, and compared by Student t or 1-factor ANOVA test as appropriate. Bicaudal 95% confidence interval (CI) was calculated as needed. Alpha error level was set at 0.05. Results From March 1st 2020 to July 31st 2022, six COVID-16 waves were identified, and 628 patients met the inclusion criteria. Overall, TnI(+) case rate was 72.0% (95% CI [68.4% - 75.4%]), and CFR was higher among TnI(+) (63.7%) than TnI(-) (32.4%; p < 0.001). As of March 1st 2020, there was a significant trend toward increasing TnI(+) case rate (p trend < 0.001) and decreasing CFR in both TnI(+) (p trend < 0.001) and TnI(-) (p trend < 0.01) across successive COVID-19 waves. Patients with TnI(+) were older (66.1 ± 14.7 years old) and had a longer TLOS (27.8 ± 24.5 days) than with TnI(-) (respectively, 59.0 ± 15.7 years old and 20.6 ± 17.0 days; p < 0.001 for both). TnI(+) case rate, TnI(+)-related and TnI(-)-related CFR, age, and TLOS across COVID-19 waves are in Table 1. Conclusion From March 1st 2020 to July 31st 2022, in patients hospitalized due to COVID-19, TnI(+) was associated with increased case fatality rate as compared to TnI(-), across six successive COVID-19 waves. Patients presenting TnI(+) were older and had longer total length-of-stay as compared to TnI(-).Table 1

Addressing the Issues of Uncertainties in Change Point Detection of Hydrological Time Series in Runoff Alteration Studies

Addressing the Issues of Uncertainties in Change Point Detection of Hydrological Time Series in Runoff Alteration Studies

Arctic marginal ice zone interannual variability and change point detection using two definitions (1983–2022)

The ongoing decline in Arctic sea ice extent and thickness underscores the scientific significance of monitoring the marginal ice zone (MIZ), a transitional region between the open ocean and pack ice. In this study, we used Bootstrap sea ice concentration (SIC) to detect the trend and change point of the Arctic MIZ over 40 years (1983–2022) using two different MIZ definitions: SIC threshold-based (MIZ t ) and SIC anomaly-based (MIZ). This study marks the exploration of a SIC anomaly-based definition of the MIZ over the Arctic. While the two MIZ definitions yield comparable seasonal trends in marginal ice zone fraction (MIZF), the MIZ fraction values peak during the transition periods (e.g. freeze-up and break-up), while the MIZ t fraction values peak in August. The analysis also uncovers consistently higher MIZF values for the MIZ than for MIZ t across all seasons. Moreover, October and August show the fastest rate of increase in MIZ t fraction and MIZ fraction, reflecting the coinciding rapid decrease in sea ice extent during those particular months. Employing the pruned exact linear time, a multiple change point detection method, highlights a significant increase in the MIZ fraction in October (after 2005) and MIZ fraction in August (after 2007). This can be indicative of the recent climate change impacts in the Arctic region that may be linked with shifts in SIC and sea ice mobility for MIZ t and MIZ, respectively.

Usage of Published Network Traffic Datasets for Anomaly and Change Point Detection

Usage of Published Network Traffic Datasets for Anomaly and Change Point Detection

Graph similarity learning for change-point detection in dynamic networks

Dynamic networks are ubiquitous for modelling sequential graph-structured data, e.g., brain connectivity, population migrations, and social networks. In this work, we consider the discrete-time framework of dynamic networks and aim at detecting change-points, i.e., abrupt changes in the structure or attributes of the graph snapshots. This task is often termed network change-point detection and has numerous applications, such as market phase discovery, fraud detection, and activity monitoring. In this work, we propose a data-driven method that can adapt to the specific network domain, and be used to detect distribution changes with no delay and in an online setting. Our algorithm is based on a siamese graph neural network, designed to learn a graph similarity function on the graph snapshots from the temporal network sequence. Without any prior knowledge on the network generative distribution and the type of change-points, our learnt similarity function allows to more effectively compare the current graph and its recent history, compared to standard graph distances or kernels. Moreover, our method can be applied to a large variety of network data, e.g., networks with edge weights or node attributes. We test our method on synthetic and real-world dynamic network data, and demonstrate that it is able to perform online network change-point detection in diverse settings. Besides, we show that it requires a shorter data history to detect changes than most existing state-of-the-art baselines.

Power-law distribution in pieces: a semi-parametric approach with change point detection

Power-law distribution in pieces: a semi-parametric approach with change point detection

A hybrid model based on discrete wavelet transform (DWT) and bidirectional recurrent neural networks for wind speed prediction

Wind speed is the main driver of wind power output, but its inherent fluctuations and deviations present significant challenges for power system security and power quality. Accurate short-term wind power forecasting is necessary to ensure the stability and seamless integration of wind energy into the grid, thereby bolstering grid reliability, optimizing power flow management, and minimizing disruptions, ultimately contributing to a more sustainable and efficient energy ecosystem. Non-stationarity is a major challenge in analyzing wind speed data, and change-point detection is essential for optimal resource allocation. This paper addresses the issue of short-term wind power forecasting for stable and effective wind energy system operation. To predict non-stationary data and detect change points, non-stationary data must first be transformed into stationary data. Discrete wavelet transform (DWT) is used to decompose wind speed traces into low- and high-frequency components for more accurate predictions using deep learning algorithms. The proposed approach utilizes a combination of DWT and recurrent neural network (RNN) models, including Bidirectional Long Short-Term Memory (BiLSTM) and Bidirectional Gated Recurrent Unit (BiGRU), complementing each other to efficiently predict short-term and long-term dependencies in wind speed data and achieve the most accurate prediction results. Experimental results substantiate the consistent superiority of the proposed approach over the baseline method, showcasing a significant enhancement in prediction accuracy, with improvements spanning from 11.36% to 16.63% for MAPE, 16.67%–26.47% for RMSE, and 21.74%–30.77% for MAE.