Adaptively constrained dynamic time warping for time series classification and clustering

Time-Normalization Approach for fNIRS Data During Tasks with High Variability in Duration

Functional near-infrared spectroscopy (fNIRS) is particularly suitable for measuring brain activity during motor tasks, due to its portability and good motion tolerance. In such cases, the trials' duration may vary depending on the experimental conditions or the participant's response, therefore a comparison of hemodynamic responses across repetitions cannot be properly performed. In this work, we present a MATLAB (R2023a) function (TaskNorm.m) developed for time-normalizing fNIRS data recorded during trials with different durations. It is based on a spline interpolation method that rescales the time -axis to the percentage of the trial with a fixed number of samples. This allows us to successively average across repetitions to obtain the mean hemodynamic responses and complete the standard data processing. The algorithm was tested on eight subjects (four with developmental coordination disorder, age: 9.78 ± 0.30 and four typically developing children, age: 9.02 ± 0.30) performing three different tasks. The results show that the TaskNorm function works as expected, allowing both a comparison and averaging of the data across multiple repetitions. The performance of the function is independent of the task or the pre-processing pipeline applied. The proposed function is publicly available and importable into the HomER3 package (v1.72.0), representing a further step in the ongoing standardization process of fNIRS data analysis.

The U-shape Association between Population Agglomeration and Individual Depression: the Role of Dialect Diversity

Depression is a relevant mental illness affecting hundreds of millions of people worldwide. As urbanization accelerates, agglomeration of populations has altered individual social network distances and life crowding, which in turn affects depressive prevalence. However, the association between depression and population agglomeration (PA) remains controversial. This study aims to explore whether and how PA could influence individual depression. Based on the China Health and Retirement Longitudinal Study (CHARLS) 2018, the empirical results showed that there was a U-shaped association between PA and individual CES-D scores. As PA increases, the risk of depression first decreases and then increases. CES-D was lowest at moderate aggregation. Dialect diversity (DD) was positively related to the incidence of individual depression. The higher the DD, the higher the risk of depression. Meanwhile, DD also played a moderating role in the association between PA and individual depression. Our observations suggest that the optimistic level of agglomeration for individual mental health is within 1500 to 2000 persons per square kilometer.

Enhancing Maritime Cybersecurity through Operational Technology Sensor Data Fusion: A Comprehensive Survey and Analysis

Cybersecurity is becoming an increasingly important aspect in ensuring maritime data protection and operational continuity. Ships, ports, surveillance and navigation systems, industrial technology, cargo, and logistics systems all contribute to a complex maritime environment with a significant cyberattack surface. To that aim, a wide range of cyberattacks in the maritime domain are possible, with the potential to infect vulnerable information and communication systems, compromising safety and security. The use of navigation and surveillance systems, which are considered as part of the maritime OT sensors, can improve maritime cyber situational awareness. This survey critically investigates whether the fusion of OT data, which are used to provide maritime situational awareness, may also improve the ability to detect cyberincidents in real time or near-real time. It includes a thorough analysis of the relevant literature, emphasizing RF but also other sensors, and data fusion approaches that can help improve maritime cybersecurity.

Non-IRC Mechanism of Bimolecular Reactions with Submerged Barriers: A Case Study of Si+ + H2O Reaction

Chemical reactions with submerged barriers may feature interesting dynamic behaviors that are distinct from those with substantial barriers or those entirely dominated by capture. The Si+ + H2O reaction is a prototypical example, involving even two submerged saddle points along the reaction path: one for the direct dissociation of H (H-dissociation SP) and another for H migration from the O-side to the Si-side (H-migration SP). We investigated the intricacies of this process by employing quasi-classical trajectory calculations on an accurate, full-dimensional ab initio potential energy surface. Through careful trajectory analysis, an interesting nonintrinsic reaction coordinate mechanism was found to play an important role in producing SiOH+ and H. This new pathway is featured as that the H atoms do not form HSiOH+ complexes along the minimum-energy path but directly dissociate into the products after passing through the H-migration SP. Furthermore, based on artificially scaled potential energy surfaces (PESs), the impact of barrier height on the reaction is also explored. This work provides new insights into the dynamics of the Si+ + H2O reaction and enriches our understanding of reactions with submerged barriers.

Linear Fuzzy Information-Granule-Based Fuzzy C-Means Algorithm for Clustering Time Series

This article aims to design a trend-oriented-granulation-based fuzzy C -means (FCM) algorithm that can cluster a group of time series at an abstract (granular) level. To achieve a better trend-oriented granulation of a time series, l1 trend filtering is firstly carried out to result in segments which are then optimized by the proposed segment merging algorithm. By constructing a linear fuzzy information granule (LFIG) on each segment, a granular time series which well reflects the linear trend characteristic of the original time series is produced. With the novel designed distance that can well measure the trend similarity of two LFIGs, the distance between two granular time series is calculated by the modified dynamic time warping (DTW) algorithm. Based on this distance, the LFIG-based FCM algorithm is developed for clustering time series. In this algorithm, cluster prototypes are iteratively updated by the specifically designed granule splitting and merging algorithm, which allows the lengths of prototypes to change in the process of iteration. This overcomes the serious drawback of the existing approaches, where the lengths of prototypes cannot be changed. Experimental studies demonstrate the superior performance of the proposed algorithm in clustering time series with different shapes or trends.

Depression prediction based on LassoNet-RNN model: A longitudinal study

Depression has become a widespread health concern today. Understanding the influencing factors can promote human mental health as well as provide a basis for exploring preventive measures. Combining LassoNet with recurrent neural network (RNN), this study constructed a screening model ,LassoNet-RNN, for identifying influencing factors of individual depression. Based on multi-wave surveys of China Health and Retirement Longitudinal Study (CHARLS) dataset (11,661 observations), we analyzed the multivariate time series data and recognized 27 characteristic variables selected from four perspectives: demographics, health-related risk factors, household economic status, and living environment. Additionally, the importance rankings of the characteristic variables were obtained. These results offered insightful recommendations for theoretical developments and practical decision making in public health.

Higher energy consumption in the evening is associated with increased odds of obesity and metabolic syndrome: findings from the 2016-2018 Korea National Health and Nutrition Examination Survey (7th KNHANES)

OBJECTIVES

Chrono-nutrition emphasizes meal timing in preventing obesity and metabolic disorders. This study explores the impact of temporal dietary patterns (TDPs) on obesity and metabolic syndrome (MetS) in Korean adults aged 20 years to 65 years.

METHODS

We utilized dynamic time warping method and Kernel k-means clustering to investigate diet quality and the odds ratios (ORs) of obesity and MetS with different TDPs using data from the 7th Korea National Health and Nutrition Examination Survey.

RESULTS

Participants were divided into three groups based on relative energy intake over 24 hours. After adjusting for age and gender, Cluster 3 (with the highest proportion of energy intake in the evening) had the lowest Healthy Eating Index scores compared to other clusters. Following adjustment for key covariates, Cluster 3 showed the highest values for body mass index, waist circumference, blood pressure, total cholesterol, and triglycerides. Compared to Cluster 1 (with a lower proportion of energy intake in the evening), Cluster 2 and Cluster 3 had ORs for obesity of 1.12 (95% confidence interval [CI], 0.97 to 1.30) and 1.19 (95% CI, 1.03 to 1.37), respectively. For MetS, the ORs were 1.26 (95% CI, 1.08 to 1.48) and 1.37 (95% CI, 1.17 to 1.61) when comparing Cluster 2 and Cluster 3 to Cluster 1.

CONCLUSIONS

This study reveals that individuals with higher energy intake in the evening have increased odds of obesity and MetS, even after adjusting for major covariates, including age and total energy intake.

Pattern-based clustering of daily weigh-in trajectories using dynamic time warping

"Smart"-scales are a new tool for frequent monitoring of weight change as well as weigh-in behavior. These scales give researchers the opportunity to discover patterns in the frequency that individuals weigh themselves over time, and how these patterns are associated with overall weight loss. Our motivating data come from an 18-month behavioral weight loss study of 55 adults classified as overweight or obese who were instructed to weigh themselves daily. Adherence to daily weigh-in routines produces a binary times series for each subject, indicating whether a participant weighed in on a given day. To characterize weigh-in by time-invariant patterns rather than overall adherence, we propose using hierarchical clustering with dynamic time warping (DTW). We perform an extensive simulation study to evaluate the performance of DTW compared to Euclidean and Jaccard distances to recover underlying patterns in adherence time series. In addition, we compare cluster performance using cluster validation indices (CVIs) under the single, average, complete, and Ward linkages and evaluate how internal and external CVIs compare for clustering binary time series. We apply conclusions from the simulation to cluster our real data and summarize observed weigh-in patterns. Our analysis finds that the adherence trajectory pattern is significantly associated with weight loss.

Automated Student Classroom Behaviors' Perception and Identification Using Motion Sensors

With the rapid development of artificial intelligence technology, the exploration and application in the field of intelligent education has become a research hotspot of increasing concern. In the actual classroom scenarios, students' classroom behavior is an important factor that directly affects their learning performance. Specifically, students with poor self-management abilities, particularly specific developmental disorders, may face educational and academic difficulties owing to physical or psychological factors. Therefore, the intelligent perception and identification of school-aged children's classroom behaviors are extremely valuable and significant. The traditional method for identifying students' classroom behavior relies on statistical surveys conducted by teachers, which incurs problems such as being time-consuming, labor-intensive, privacy-violating, and an inaccurate manual intervention. To address the above-mentioned issues, we constructed a motion sensor-based intelligent system to realize the perception and identification of classroom behavior in the current study. For the acquired sensor signal, we proposed a Voting-Based Dynamic Time Warping algorithm (VB-DTW) in which a voting mechanism is used to compare the similarities between adjacent clips and extract valid action segments. Subsequent experiments have verified that effective signal segments can help improve the accuracy of behavior identification. Furthermore, upon combining with the classroom motion data acquisition system, through the powerful feature extraction ability of the deep learning algorithms, the effectiveness and feasibility are verified from the perspectives of the dimensional signal characteristics and time series separately so as to realize the accurate, non-invasive and intelligent children's behavior detection. To verify the feasibility of the proposed method, a self-constructed dataset (SCB-13) was collected. Thirteen participants were invited to perform 14 common class behaviors, wearing motion sensors whose data were recorded by a program. In SCB-13, the proposed method achieved 100% identification accuracy. Based on the proposed algorithms, it is possible to provide immediate feedback on students' classroom performance and help them improve their learning performance while providing an essential reference basis and data support for constructing an intelligent digital education platform.

Functional distributional clustering using spatio-temporal data

This paper presents a new method called the functional distributional clustering algorithm (FDCA) that seeks to identify spatially contiguous clusters and incorporate changes in temporal patterns across overcrowded networks. This method is motivated by a graph-based network composed of sensors arranged over space where recorded observations for each sensor represent a multi-modal distribution. The proposed method is fully non-parametric and generates clusters within an agglomerative hierarchical clustering approach based on a measure of distance that defines a cumulative distribution function over temporal changes for different locations in space. Traditional hierarchical clustering algorithms that are spatially adapted do not typically accommodate the temporal characteristics of the underlying data. The effectiveness of the FDCA is illustrated using an application to both empirical and simulated data from about 400 sensors in a 2.5 square miles network area in downtown San Francisco, California. The results demonstrate the superior ability of the the FDCA in identifying true clusters compared to functional only and distributional only algorithms and similar performance to a model-based clustering algorithm.

Applications of monitoring and tracing the evolution of clustering solutions in dynamic datasets

The clustering approach is widely accepted as the most prominent unsupervised learning problem in data mining techniques. This procedure deals with the identification of notable structures in unlabeled datasets. In modern days clustering of dynamic data, streams play a vital role in policy-making, and researchers are paying particular attention to monitoring the evolution of clustering solutions over time. The data streams evolve continually, and different sources generate data items over time. The clustering solution over this stream is not stationary and changes with the influx of new data items. This paper presents a comprehensive study of algorithms related to tracing the evolution of clusters over time in cumulative datasets. To demonstrate the applications and significance of the tracing cluster evolution, we implement the MONIC algorithm in R-software. This article illustrates how the data segmentation of dynamic streams is done and shows the applications of monitoring changes in clustering solutions with the help of real-life published datasets.

clusTransition: An R package for monitoring transition in cluster solutions of temporal datasets

Clustering analysis' primary purpose is to divide a dataset into a finite number of segments based on the similarities between items. In recent years, a significant amount of study has focused on the spatio-temporal aspects of clustering. However, clusters are no longer regarded as static objects since changes influence them in the underlying population. This paper describes an R package implementing the MONIC framework for tracing the evolution of clusters extracted from temporal datasets. The name of the package is clusTransition, which stands for Cluster Transition. The algorithm is based on re-clustering cumulative datasets that evolve at successive time-points and monitoring the transitions experienced by the clusters in these clustering solutions. This paper's contribution is to demonstrate how the package clusTransition is developed in the R programming language, and its workflow is discussed using hypothetical and real-life datasets.

Two-dimensional dynamic time warping algorithm for matrices similarity

Dynamic Time Warping (DTW algorithm) provides an effective method to obtain the similarity between unequal-sized signals. However, it cannot directly deal with high-dimensional samples such as matrices. Expanding a matrix to one dimensional vector as the input data of DTW will decrease the measure accuracy because of the losing of position information in the matrix. Aiming at this problem, a two-dimensional dynamic time warping algorithm (2D-DTW) is proposed in this paper to directly measure the similarity between matrices. In 2D-DTW algorithm, a three dimensional distance-cuboid is constructed, and its mapped distance matrix is defined by cutting and compressing the distance-cuboid. By introducing the dynamic programming theory to search the shortest warping path in the mapped matrix, the corresponding shortest distance can be obtained as the expected similarity measure. The experimental results suggest that the performance of 2D-DTW distance is superior to the traditional Euclidean distance and can improve the similarity accuracy between matrices by introducing the warping alignment mechanisms. 2D-DTW algorithm extends the application ranges of traditional DTW and is especially suitable for high-dimensional data.

Optimization-Assisting Dual-Step Clustering of Time Series Data

This paper aims to propose a new time series data clustering with the following steps: (1) data reduction and (2) clustering. The main objective of the time series data clustering is to minimize the dataset size via a prototype defined for same time series data in every group that significantly reduced the complexities. Initially, the time series dataset in the data reduction step is subjected to preprocessing process. Further, in the proposed probability based distance measure evaluation, the time series data is grouped into subclusters. In the clustering step, the proposed shape based similarity measure is performed. Moreover, the clustering process is carried out by optimized k-mean clustering in which the center point is optimally tuned by a new customized whale optimization algorithm (CWOA). At last, the performance of the adopted model is computed to other traditional models with respect to various measures such as sensitivity, accuracy, FPR, conentropy, precision, FNR, specificity, MCC, entropy, F-measure, and Rand index, respectively.

Is the Spatial-Temporal Dependence Model Reliable for the Short-Term Freight Volume Forecast of Inland Ports? A Case Study of the Yangtze River, China

The purpose of this study is to investigate whether spatial-temporal dependence models can improve the prediction performance of short-term freight volume forecasts in inland ports. To evaluate the effectiveness of spatial-temporal dependence forecasting, the basic time series forecasting models for use in our comparison were first built based on an autoregression integrated moving average model (ARIMA), a back-propagation neural network (BPNN), and support vector regression (SVR). Subsequently, combining a gradient boosting decision tree (GBDT) with SVR, an SVR-GBDT model for spatial-temporal dependence forecast was constructed. The SVR model was only used to build a spatial-temporal dependence forecasting model, which does not distinguish spatial and temporal information but instead takes them as data features. Taking inland ports in the Yangtze River as an example, the results indicated that the ports’ weekly freight volumes had a higher autocorrelation with the previous 1–3 weeks, and the Pearson correlation values of the ports’ weekly cargo volume were mainly located in the interval (0.2–0.5). In addition, the weekly freight volumes of the inland ports were higher depending on their past data, and the spatial-temporal dependence model improved the performance of the weekly freight volume forecasts for the inland river. This study may help to (1) reveal the significance of spatial correlation factors in ports’ short-term freight volume predictions, (2) develop prediction models for inland ports, and (3) improve the planning and operation of port entities.

A Fuzzy Radial Basis Adaptive Inference Network and Its Application to Time-Varying Signal Classification

A fuzzy radial basis adaptive inference network (FRBAIN) is proposed for multichannel time-varying signal fusion analysis and feature knowledge embedding. The model which combines the prior signal feature embedding mechanism of the radial basis kernel function with the rule-based logic inference ability of fuzzy system is composed of a multichannel time-varying signal input layer, a radial basis fuzzification layer, a rule layer, a regularization layer, and a T-S fuzzy classifier layer. The dynamic fuzzy clustering algorithm was used to divide the sample set pattern class into several subclasses with similar features. The fuzzy radial basis neurons (FRBNs) were defined and used as parameterized membership functions, and typical feature samples of each pattern subclass were used as kernel centers of the FRBN to realize the embedding of the diverse prior feature knowledge and the fuzzification of the input signals. According to the signal categories of FRBN kernel centers, nodes in the rule layer were selectively connected with nodes in the FRBN layer. A fuzzy multiplication operation was used to achieve synthesis of pattern class membership information and establishment of fuzzy inference rules. The excitation intensity of each rule was used as the input of T-S fuzzy classifier to classify the input signals. The FRBAIN can adaptively establish fuzzy set membership functions, fuzzy inference, and classification rules based on the learning of sample set, realize structural and data constraints of the model, and improve the modeling properties of imbalanced datasets. In this paper, the properties of FRBAIN were analyzed and a comprehensive learning algorithm was established. Experimental validation was performed with classification diagnoses from four complex cardiovascular diseases based on 12-lead ECG signals. Results demonstrated that, in the case of small-scale imbalanced datasets, the proposed method significantly improved both classification accuracy and generalizability comparing with other methods in the experiment.