Sequential Model Research Articles

How Ethical Leadership Fosters the Development of Public University Employees’ Social Capital in Vietnam? A Sequential Mediation Model

How Ethical Leadership Fosters the Development of Public University Employees’ Social Capital in Vietnam? A Sequential Mediation Model

Sequential Multi-Scale Modeling Using an Artificial Neural Network-Based Surrogate Material Model for Predicting the Mechanical Behavior of a Li-Ion Pouch Cell Under Abuse Conditions

To analyze the safety behavior of electric vehicles, mechanical simulation models of their battery cells are essential. To ensure computational efficiency, the heterogeneous cell structure is represented by homogenized material models. The required parameters are calibrated against several characteristic cell experiments. As a result, it is hardly possible to describe the behavior of the individual battery components, which reduces the level of detail. In this work, a new data-driven material model is presented, which not only provides the homogenized behavior but also information about the components. For this purpose, a representative volume element (RVE) of the cell structure is created. To determine the constitutive material models of the individual components, different characterization tests are performed. A novel method for carrying out single-layer compression tests is presented for the characterization in the thickness direction. The parameterized RVE is subjected to a large number of load cases using first-order homogenization theory. This data basis is used to train an artificial neural network (ANN), which is then implemented in commercial FEA software LS-DYNA R9.3.1 and is thus available as a material model. This novel data-driven material model not only provides the stress–strain relationship, but also outputs information about the condition of the components, such as the thinning of the separator. The material model is validated against two characteristic cell experiments. A three-point-bending test and an indentation test of the cell is used for this purpose. Finally, the influence of the architecture of the neural network on the computational effort is discussed.

Aggregation unveiled: A sequential modelling approach to bark beetle outbreaks

Tree-killing bark beetle infestations are a cause of massive coniferous forest mortality impacting forest ecosystems and the ecosystem services they provide. Models predicting bark beetle outbreaks are crucial for forest management and conservation, necessitating studies of the effect of epidemiological traits on the probability and severity of outbreaks. Due to the aggregation behaviour of beetles and host tree defence, this epidemiological interaction is highly non-linear and outbreak behaviour remains poorly understood, motivating questions about when an outbreak can occur, what determines outbreak severity, and how aggregation behaviour modulates these quantities. Here, we apply the principle of distributed delays to create a novel and mathematically tractable model for beetle aggregation in an epidemiological framework. We derive the critical outbreak threshold for the beetle emergence rate, which is a quantity analogous to the basic reproductive ratio, R0, for epidemics. Beetle aggregation qualitatively impacts outbreak potential from depending on the emergence rate alone in the absence of aggregation to depending on both emergence rate and initial beetle density when aggregation is required. Finally, we use a stochastic model to confirm that our deterministic model predictions are robust in finite populations.

Advanced Sequential Modeling for DeepFake Audio Identification

Advanced Sequential Modeling for DeepFake Audio Identification

Digital twins as a tool for shaping the business environment of modern organizations

Subject. The article is devoted to the technology of digital twins as a tool for the formation of the business environment of modern organizations. Objectives. The aim is to show the need to improve the quality of business environment of modern organizations and improve the efficiency of business process management, to justify the importance of using modern digital technologies, to explore the technology of digital twins as a promising tool for shaping the business environment of modern organizations. Methods. The study rests on systems approach enabling to consider the technology of digital twins as a set of sequential modeling processes, which results in the creation of an adequate digital virtual reality. I also use the content analysis of available sources and statistical data. Results. The use of digital twins technology provides real opportunities for enterprises to operate in a dynamic market environment. However, their implementation faces a number of challenges that must be considered when deciding to implement the technology under study in the practical activities of Russian enterprises. Moreover, a fairly long investment and project cycle does not end with the creation of a digital twin, but requires constant work related to its improvement, saturation with information, technical and technological maintenance of this system. Conclusions. Currently, an urgent task is to develop pilot projects for the creation of digital twins implemented at experimental sites. These projects will provide the necessary information on the feasibility and specifics of this system implementation, and help gain necessary experience in its development, implementation, and operation.

Simultaneous Vibration and Temperature Real-Time Monitoring Using Single Fiber Bragg Grating and Free Space Optics

Simultaneous monitoring of temperature and vibration of electromechanical systems, public buildings, and volcanic terrain is an important indicator of preventing accidents. This work showcases the simultaneous temperature and vibration measurement using a single-fiber Bragg grating (FBG) sensor. The concurrent interrogation of vibration and temperature by the FBG sensing system can be integrated with free space optics (FSO), which saves on the costs associated with fiber optic cables and overcomes terrain barriers. Furthermore, the real-time measured vibration frequencies and different temperature levels can be recognized by a deep neural network (DNN) sequential model. In addition to the FSO support, the sensing system can be arranged in multiple sensing locations to form a wide range of measurements, offering a highly cost-effective solution.

Explainable next POI recommendation based on spatial–temporal disentanglement representation and pseudo profile generation

The current research in Point-of-Interest (POI) recommendation primarily aims to decipher users’ transitional patterns to predict their future location visits. Traditional approaches often intertwine various features to model these check-in transitions, which inadvertently compromises the quality of the resulting representations. This issue is compounded in both industrial and academic settings, where user-generated textual data is frequently inaccessible or restricted due to privacy concerns. Such limitations in user profiles pose significant challenges to the effectiveness of subsequent applications. In response to these challenges, the recent rise of Large Language Models (LLMs) offers a novel perspective. Diverging from the conventional approach of leveraging LLMs for semantic-based next check-in predictions, our research investigates the potential of integrating LLMs with sequential recommendation systems. This integration aims to augment feature dimensions and facilitate the generation of explicit explanations. To this end, we introduce CrossDR-Gen, a Cross-sequence Location Disentanglement Representation methodology. CrossDR-Gen is specifically designed for next POI recommendation and explanation generation. It uniquely considers spatial and temporal factors in shaping check-in behaviors, offering a comprehensive global view of location transitions. Crucially, CrossDR-Gen utilizes LLMs for pseudo profile generation in scenarios with limited semantic context, thereby enriching user features without relying on additional textual profiles or conversational data. Our experiments on real-world datasets demonstrate that CrossDR-Gen not only excels in addressing cold-start scenarios but also showcases robust recommendation capabilities. These findings validate the effectiveness of our proposed cooperative paradigm between LLMs and sequential recommendation models, highlighting a promising avenue for future research in POI recommendation systems.

Let long-term interests talk: An disentangled learning model for recommendation based on short-term interests generation

In e-commerce recommendation systems, users’ long-term and short-term interests jointly influence product selection. However, the behavioral conformity phenomenon tends to be more prominent in short-term sequences, and the entanglement of true preference and popularity conformity data confuses the user’s real interest needs. To address this issue, we propose a sequential recommendation model called DFRec to disentangle short-term interests from popularity bias. By leveraging long-term interest trends, the model promotes the separation of short-term interests from popularity-driven deviations, thereby reducing the impact of popularity interference in short-term sequences. Firstly, we propose a Disentangled Frequency Attention Network(DFAN) to address the entanglement between real sequence features and conformity data in users’ short-term behavioral sequences. The approach clarify the non-entangled representation of the user’s short-term interest and conformity on the basis of long-term interest trends. Secondly, in order to capture the real long-term interest characteristics of users, this paper suggests using a Learnable Filter(LF) to filter the noise frequencies in long-term sequence. The method decouples the horizontal and vertical directions of the sequence and filters out the noise in both directions. Finally, consider the importance of the two interests characteristics is dynamic, we propose a joint learning framework with dual embeddings to balance and fusion these two features of users’ interests. Experimental results on three public datasets demonstrate that our model effectively captures dynamic user interests and outperforms six baseline models.

Brain Tumour Detection Using Deep Learning and Image Processing

Abstract —Brain tumor detection is a critical aspect of neurooncology, where timely and accurate diagnosis significantly impacts patient outcomes. Traditional imaging techniques, while useful, are often limited by variations in image quality, illumination inconsistencies, and the complex nature of brain tumors themselves, which vary widely in size, shape, and texture. This paper presents a deep learning- based approach to brain tumor detection using Convolutional Neural Networks (CNNs) combined with advanced image processing methods. Our methodology involves preprocessing brain MRI scans using histogram equalization, morphological operations, and data augmentation to address illumination issues and enhance the tumor region. Following preprocessing, the images are fed into a CNN, where transfer learning from the Sequential model is applied to improve classification accuracy. Experimental results demonstrate high recall and precision rates, affirming the model’s robustness in detecting brain tumors in MRI images. This approach not only addresses the challenge of inconsistent MRI scan quality but also reduces the risk of misclassification, showing promise for clinical application. Keywords— Brain Tumor Detection, Computer-aided Diagnosis, Computer Vision, Convolutional Neural Networks, Deep Learning, Image Processing, Transfer Learning.

Buyer’s market-oriented culture, strategic procurement capability, and customer performance in the public procurement of innovations

Abstract The public procurement of innovations is a demand-side policy instrument that harnesses supplier innovations to improve products and services delivered to the customers. This study analyses a nation-wide survey of Finnish public sector contracting authorities. A sequential mediation model suggests that the public sector buyer’s market-oriented culture is positively associated with the procurer’s strategic procurement capability, which in turn predicts the use of public procurement of innovations to create customer value for customers. These results are robust to a potentially biased perception of customer performance.

Reputation vs. price: Sequential recommendations based on cue diagnosticity theory

Sequential recommendations have been widely used in e-commerce platforms to effectively capture consumers' dynamic preferences and provide them with preferred products. Traditional models usually use ratings and product attributes for sequential recommendations to satisfy consumers’ more personalized needs. Consumers also rely on reviews from other consumers to form a general impression of the product or retailer before making their purchase decisions. Such impressions can be treated as reputations of the product or retailer. Inspired by cue diagnosticity theory, we divide the attributes related to product purchase into low- and high-scope cues. High-scope cues, including reputations, are not easily changed because they are formed over a long period by numerous consumers, whereas low-scope cues, such as price, can be easily changed by retailers. We propose an innovative Sequential Recommendation model by Integrating Low-scope cues and High-scope cues (SRILH). We design a cue-extraction layer to extract high-scope cues from consumer online reviews and a hierarchical cue-aware attention layer to learn the joint effect of low- and high-scope cues. We evaluate the performance of the proposed model using three real-world datasets, and our experimental results validate its effectiveness and robustness. Our research contributes to sequential recommendations research by uncovering the joint effects of cues on consumer behavior and by providing valuable insights into the dynamics of cue preference formation in recommendation systems. We also extend the empirical literature on cue diagnosticity theory by drawing conclusions from the micro and individual perspectives to shed light on how different cues impact consumer choices. The interpretable visualization results provide managerial insights for retailers and manufacturers to improve their products.

Graph-based rank aggregation: a deep-learning approach

Purpose This study aims to introduce a novel rank aggregation algorithm that leverages graph theory and deep-learning to improve the accuracy and relevance of aggregated rankings in metasearch scenarios, particularly when faced with inconsistent and low-quality rank lists. By strategically selecting a subset of base rankers, the algorithm enhances the quality of the aggregated ranking while using only a subset of base rankers. Design/methodology/approach The proposed algorithm leverages a graph-based model to represent the interrelationships between base rankers. By applying Spectral clustering, the algorithm identifies a subset of top-performing base rankers based on their retrieval effectiveness. These selected rankers are then integrated into a sequential deep-learning model to estimate relevance labels for query-document pairs. Findings Empirical evaluation on the MQ2007-agg and MQ2008-agg data sets demonstrates the substantial performance gains achieved by the proposed algorithm compared to baseline methods, with an average improvement of 8.7% in MAP and 11.9% in NDCG@1. The algorithm’s effectiveness can be attributed to its ability to effectively integrate diverse perspectives from base rankers and capture complex relationships within the data. Originality/value This research presents a novel approach to rank aggregation that integrates graph theory and deep-learning. The author proposes a graph-based model to select the most effective subset for metasearch applications by constructing a similarity graph of base rankers. This innovative method addresses the challenges posed by inconsistent and low-quality rank lists, offering a unique solution to the problem.

Data-driven mechanical behavior modeling of granular biomass materials

Significant equipment downtime, mainly caused by the variable attributes and complex mechanical behavior of granular biomass materials, has challenged the great potential of biofuel generation. Expensive and time-consuming lab experiments are not affordable in fully characterizing material behavior for samples with all possible attributes. In this paper, we present data-driven modeling of the cyclic axial compression and ring-shear stress–strain behavior for granular biomass materials with different moisture contents and mean particle sizes. Laboratory characterization data of milled pine samples, with a mean particle size of 2, 4, and 6 mm and a moisture content of 0%, 20%, and 40%, were employed to train two machine learning (ML) models independently. The sequential model takes a sequence of loading history into Gated Recurrent Units (GRU) and predicts the next step of strain/stress with a feed-forward neural network. In contrast, the incremental model takes only the current state to predict the stiffness/compliance coefficient and compute the stress/strain increment. Both models account for material variations in particle sizes and moisture content. After training, both models demonstrate their high efficiency and accuracy in predicting the mechanical behavior of samples with a diverse set of unseen material properties, augmenting the existing laboratory data set. They also effectively capture the underlying physics, which involves increased compressibility as moisture content rises and a linear relationship between applied compression and shear strength. This effort established the foundation for comprehensive data-driven constitutive modeling tailored for unconventional granular materials.

An efficient sequential Kriging model for structure safety lifetime analysis considering uncertain degradation

Safety lifetime analysis performs a crucial role in ensuring structural safety in service and developing effective maintenance strategies, which also places higher demands on calculation. However, existing safety lifetime analysis methods generally suffer from inefficiency, which is more prominent for complex engineering structures. In this paper, a novel sequential single-loop Kriging (SSK) surrogate modeling approach is proposed to calculate the safety lifetime in an efficient and accurate manner. To reduce the computational cost, a single-loop safety lifetime analysis framework is proposed. In this framework, there is no need to accurately calculate the time-dependent failure probability (TDFP) in any sub-time interval. By searching the safety lifetime in the process of time-dependent reliability analysis (TRA) and dynamically adjusting the interest time interval, the safety lifetime can be quickly determined by constructing only one Kriging model. To maximize the utilization of sample information, SSK employs a modified learning function that allows most of the training points to be added before the safety lifetime. For accuracy, a convergence criterion that includes two Kriging models is proposed. Mathematical engineering examples are used to illustrate the accuracy and efficiency of SSK. The proposed method offers a promising approach for efficient safety lifetime analysis of engineering problems.

A random sequential adsorption model for the prediction of SARS-CoV-2 spike protein spatial and orientational distribution on a sensing surface

The interaction between antibodies and viral proteins is crucial for immune response, and understanding these interactions is therefore of great importance. The sensitivity of immunosensors, where antigens are immobilized, largely depends on their orientation. However, experimental methods to determine the orientation are often time-consuming and require specialized equipment. In this paper, we present a random sequential adsorption (RSA) model for the spatial and orientational distribution of the SARS-CoV-2 spike protein immobilized on an 11-mercaptoundecanoic acid self-assembling monolayer. We compare the RSA model prediction with experimentally obtained results for the surface mass density, 581±28 fmol/cm2, and the average number of available RBD per spike protein, 1.19. Experimental results were in good agreement for both immobilized SCoV2-S monolayer (560 fmol/cm2) and SCoV2-S/mAb (634 fmol/cm2). This validation of our simulation results allows us to draw conclusions about binding site density and the observed high sensitivity of the immunosensors. Our findings provide important insights into epitope density and immunosensor sensitivity, offering substantial utility for advancing biosensor research methodologies.

Sequential model of FSN classification with ZABLS slotting and vehicle routing problem using hybridization of ant colony optimization and tabu search to reduce picking time

Purpose: The research purpose is to enhance picking performance by developing a hybrid algorithm that classifies SKU, determines slot, and routes the process for Delivery Order (DO).Design/methodology/approach: FSN classification is used for categorizing the products into three groups: fast-moving, slow-moving, and not-moving based on the consumption rate and average stay of each SKU. The result of classification is continued with ZABLS Slotting for product placement on warehouse shelves based on the quickest to the longest picking time on each slot. Slotting results are used for VRP addresses with Ant Colony Optimization and Tabu Search hybridization algorithms.Due to the high process and cost to move the goods storage location because of the use of FSN and ZABLS method, hybridization algorithms are compared to pre-slotting and post-slotting conditions. Method verification uses 30 random sampling DO, and based on existing storage locations, it took 757.14 seconds average picking time.Findings: On pre-slotting condition, it reduced 17.74% to 626.34 seconds, and on post-slotting condition it reduced 25.75% to 557.64 seconds.The reduced picking time gives PT. XYZ better performance on fulfilling delivery orders in a day; theoretically, based on standard time, PT. XYZ can fulfill 40 orders in a day, and based on current performance, PT. XYZ can only fulfill 31 DO in a day. The uses of ACO-TS hybridization algorithms on pre-slotting condition PT. XYZ can fulfill 45 DO in a day and on post-slotting condition PT. XYZ can fulfill 51 DO in a day, increasing 27.5% from current performance.Originality/value: The novelty of this research is the use of hybridization algorithms of Ant Colony Optimization and Tabu Search (ACO-TS) to design sequential model of FSN-ZABLS to VRP to minimize picking time on each Delivery Order (DO).

Dangerous depths of bifurcation: the rise of “inter-national security narcissists” and undersea cable (dis) connections

ABSTRACT This article examines the rising bifurcation in global undersea cable networks amid growing U.S.-China tensions, building on the “weaponized interdependence” (WI) framework. While WI sheds light on how states leverage network centrality for strategic benefits, it falls short of capturing the proactive reshaping of network topologies for national security. We argue this bifurcation fosters parallel, redundant systems that, despite increased costs, enhance control and security. A sequential game model illustrates how U.S. and Chinese security-driven decisions contribute to network divergence, while a hierarchical Q-learning model explores how path dependency and lock-in effects sustain this split. Empirical cases, like the Cap-1 project, demonstrate the practical impact, emphasizing security’s role in reconfiguring global communications infrastructure.

Exploring the Impact of Word2Vec Embeddings Across Neural Network Architectures for Sentiment Analysis

Abstract. Sentiment analysis is crucial for understanding public opinion, gauging customer satisfaction, and making informed business decisions based on the emotional tone of textual data. This study investigates the performance of different Word2Vec-based embedding strategies static, non-static, and multichannel for sentiment analysis across various neural network architectures, including Convolution Neural Networks (CNNs), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRUs). Despite the rise of advanced contextual embedding methods such as Bidirectional Encoder Representations from Transformers (BERT), Word to Vector (Word2Vec) retains its importance due to its simplicity and lower computational demands, making it ideal for use in settings with limited resources. The goal is to evaluate the impact of fine-tuning Word2Vec embeddings on the accuracy of sentiment classification. Using the Internet Movie Database (IMDb), this work finds that multichannel embeddings, which combine static and non-static representations, provide the best performance across most architectures, while static embeddings continue to deliver strong results in specific sequential models. These findings highlight the balance between efficiency and accuracy in traditional word embeddings, particularly when advanced models are not feasible.

Exploring the Impact of Word2Vec Embeddings Across Neural Network Architectures for Sentiment Analysis

Abstract. Sentiment analysis is crucial for understanding public opinion, gauging customer satisfaction, and making informed business decisions based on the emotional tone of textual data. This study investigates the performance of different Word2Vec-based embedding strategies static, non-static, and multichannel for sentiment analysis across various neural network architectures, including Convolution Neural Networks (CNNs), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRUs). Despite the rise of advanced contextual embedding methods such as Bidirectional Encoder Representations from Transformers (BERT), Word to Vector (Word2Vec) retains its importance due to its simplicity and lower computational demands, making it ideal for use in settings with limited resources. The goal is to evaluate the impact of fine-tuning Word2Vec embeddings on the accuracy of sentiment classification. Using the Internet Movie Database (IMDb), this work finds that multichannel embeddings, which combine static and non-static representations, provide the best performance across most architectures, while static embeddings continue to deliver strong results in specific sequential models. These findings highlight the balance between efficiency and accuracy in traditional word embeddings, particularly when advanced models are not feasible.

Deep Learning-Enabled De-Noising of Fiber Bragg Grating-Based Glucose Sensor: Improving Sensing Accuracy of Experimental Data

This paper outlines the successful utilization of deep learning (DL) techniques to elevate data quality for assessing Au-TFBG (tilted fiber Bragg grating) sensor performance. Our approach involves a well-structured DL-assisted framework integrating a hierarchical composite attention mechanism. In order to mitigate high variability in experimental data, we initially employ seasonal decomposition using moving averages (SDMA) statistical models to filter out redundant data points. Subsequently, sequential DL models extrapolate the normalized transmittance (Tn) vs. wavelength spectra, which showcases promising results through our SpecExLSTM model. Furthermore, we introduce the AttentiveSpecExLSTM model, integrating a composite attention mechanism to improve Tn sequence prediction accuracy. Evaluation metrics demonstrate its superior performance, including a root mean square error of 1.73 ± 0.05, a mean absolute error of 1.20 ± 0.04, and a symmetric mean absolute percentage error of 2.22 ± 0.05, among others. Additionally, our novel minima difference (Min. Dif.) metric achieves a value of 1.08 ± 0.46, quantifying wavelength for the global minima within the Tn sequence. The composite attention mechanism in the AttentiveSpecExLSTM adeptly captures both high-level and low-level dependencies, refining the model’s comprehension and guiding informed decisions. Hierarchical dot and additive attention within this model enable nuanced attention refinement across model layers; dot attention focuses on high-level dependencies, while additive attention fine-tunes its focus on low-level dependencies within the sequence. This innovative strategy enables accurate estimation of the spectral width (full-width half maxima) of the Tn curve, surpassing raw data’s capabilities. These findings significantly contribute to data quality enhancement and sensor performance analysis. Insights from this study hold promise for future sensor applications, enhancing sensitivity and accuracy by improving experimental data quality and sensor performance assessment.