Search

Post-earthquake safety assessment of buildings and infrastructure poses significant challenges, often relying on time-consuming visual inspections. To expedite this process, safety criteria based on a demand-capacity model are utilized. However, rapid assessment frameworks require accurate estimations of intensity measures (IMs) to estimate seismic demand and assess structural health. Unfortunately, post-earthquake IM values are typically only available at monitored locations equipped with sensors or monitoring systems, limiting broader assessments. Simple spatial interpolation methods, while possible, struggle to consider crucial physical factors such as earthquake magnitude, epicentral distance, and soil type, leading to substantial estimation errors, especially in areas with insufficient or non-uniform seismic station coverage. To address these issues, a novel framework, BN-GMPE, combining a Bayesian network (BN) and a ground motion prediction equation (GMPE), is proposed. BN-GMPE enables inference and prediction under uncertainty, incorporating physical parameters in seismic wave propagation. A further novelty introduced in this work regards separating the near and far seismic fields in the updating process to attain a clearer understanding of uncertainty and more accurate IM estimation. In the proposed approach, a GMPE is employed for the estimation, and the bias and standard deviation of the prediction error are updated after any new information is entered into the network. The proposed method is benchmarked against a classic Kriging interpolator technique, considering some recent earthquake shocks in Italy. The proposed BN framework can naturally extend for estimating the probability of failure of various structures in a targeted region, which represents the ultimate aim of this research. [ABSTRACT FROM AUTHOR]

Globally, COVID-19 had devastating consequences on the construction sector, but there is a little knowledge of how the pandemic impacted developing nations. This study aims to investigate the job stability of employees in the construction sector and to highlight the factors that negatively affect employment. A survey of 1000 questionnaires was distributed to construction sector personnel and 436 valid responses were returned. Three popular data mining techniques were utilized: binary logistic regression, support vector machine, and Bayesian networks. Twelve models were developed; one binary logistic model, four support vector machine models, and six Bayesian networks models. The results of these models were compared based on accuracy, sensitivity, specificity, precision, recall, F-measure, and ROC Area. As a result, it was found that the method of the Bayesian networks was more effective in modelling the job stability of employees in comparison with the other models. According to our study, craft labourers were most affected in terms of job losses, followed by site engineers, while project managers and contractors were the least affected. The findings highlight the importance of protecting the most vulnerable labourers by revising the current legislation policy initiatives that should concentrate on establishing a better and more sustainable labour market. [ABSTRACT FROM AUTHOR]

In risk assessment, numerous subfactors influence the probabilities of the main factors. These main factors reflect adverse outcomes, which are essential in risk assessment. A Bayesian network can model the entire set of subfactors and their interconnections. To assess the probabilities of all possible states of the main factors (adverse consequences), complete information about the probabilities of all relevant subfactor states in the network nodes must be utilized. This is a typical task of probabilistic inference. The algorithm proposed by J. Pearl is widely used for point estimates of relevant probabilities. However, in many practical problems, including environmental risk assessment, it is not possible to assign crisp probabilities for relevant events due to the lack of sufficient statistical data. In such situations, expert assignment of probabilities is widely used. Uncertainty in expert assessments can be successfully modeled using triangular fuzzy numbers. That is why this article proposes a fuzzy version of this algorithm, which can solve the problem of probabilistic inference on a Bayesian network when the initial probability values are given as triangular fuzzy numbers. [ABSTRACT FROM AUTHOR]

The relationship between metabolically healthy obesity (MHO) and cardiovascular disease (CVD) risk remains debated. The critical point may be the lack of consensus on MHO’s definition and diagnostic criteria. This study aimed to investigate the association of MHO status with arteriosclerosis-CVD (ASCVD) risk in Chinese under new diagnostic criteria. Participants aged 35–79 in the 2009 China Health and Nutrition Survey cohort were included. The 10-year ASCVD risk was predicted by the prediction for ASCVD risk in China, and participants with a predicted risk of ⩾ 10% were classified into the high-risk group. The Bayesian network (BN) models were constructed to characterize the multivariable probabilistic connections between metabolically obesity phenotypes and ASCVD risk. The 10-year ASCVD risk score and the proportion of individuals at ASCVD high risk were significantly different between metabolically obesity phenotypes (P< 0.001). BN reasoning results showed that MHO individuals were not significantly associated with a 10-year ASCVD risk. Among metabolically unhealthy individuals, the conditional probability of high ASCVD risk increased with the Body Mass Index (BMI), with the conditional probability of high ASCVD risk was 24.63% (95% CI: 22.81–26.55%), 32.97% (95% CI: 30.75–35.27%) and 40.2% (95% CI: 36.64–43.86%) for metabolically unhealthy normal weight (MUNW), metabolically healthy overweight weight (MHOW), and metabolically unhealthy obesity (MUO) group, respectively. Subgroup analysis showed that MHO individuals were at increased risk of CVD compared with metabolically healthy normal weight (MHNW) individuals only in females. These results showed that there was no significant increase in ASCVD risk of MHO phenotype based on the new diagnostic criteria, suggesting that MHO is in a relatively healthy state. [ABSTRACT FROM AUTHOR]

Energy policy has a significant impact on the state of the environment and, therefore, on residents' health and life expectancy, especially in highly urbanized areas. Reducing emissions is currently one of the necessary actions that must be taken at the scale of individual countries to ensure sustainable development. The article aims to identify the best ways to shape energy policy by evaluating development scenarios for air protection and their environmental impact. The realization of the goal is based on the data included in three groups: (1) Economic factors, Health factors, and Demographic factors; (2) Clima-e related economic losses, Renewable Energy sources in electricity, heating, and cooling, Premature deaths due to exposure to fine particulate matter (PM2.5), Health impacts of air pollution, Population change; (3) Demographic balance and crude rates at the national level, GDP per capita in purchasing power PPS, GDP, and principal components; covering 36 EU countries in 2019 and 2021. The study proposes an advanced methodology for assessing development strategies by integrating the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) and Bayesian networks (BN) and incorporating them into a multicriteria decision-making (MCDM) support system. The TOPSIS model based on BN allowed for the illustration of the features of many criteria and the identification of relationships between scenarios, allowing for selecting the best way to develop energy policy. The results showed a 60.39% chance of achieving success in extending the life of residents by five years. At the same time, the most favorable development path was the scenario promoting activities aimed at reducing air pollution by introducing renewable energy sources to produce energy used for lighting and preparing domestic hot water urban areas. By presenting possible scenarios and the probability of success, it is possible to achieve the goal of practical energy policy at the level of the country and individual European cities and also by extending the life of city inhabitants, as presented by the authors in this study. [ABSTRACT FROM AUTHOR]

SecuriDN v. 0.1 is a tool for the representation of the assets composing the IT and the OT subsystems of Distributed Energy Resources (DERs) control networks and the possible cyberattacks that can threaten them. It is part of a platform that allows the evaluation of the security risks of DER control systems. SecuriDN is a multi-formalism tool, meaning that it manages several types of models: architecture graph, attack graphs and Dynamic Bayesian Networks (DBNs). In particular, each asset in the architecture is characterized by an attack graph showing the combinations of attack techniques that may affect the asset. By merging the attack graphs according to the asset associations in the architecture, a DBN is generated. Then, the evidence-based and time-driven probabilistic analysis of the DBN permits the quantification of the system security level. Indeed, the DBN probabilistic graphical model can be analyzed through inference algorithms, suitable for forward and backward assessment of the system's belief state. In this paper, the features and the main goals of SecuriDN are described and illustrated through a simplified but realistic case study. [ABSTRACT FROM AUTHOR]

A System-of-Systems (SoS) is characterized both by independence and by inter-dependency. This inter-dependency, while allowing an SoS to achieve its objectives, also means that failures can cascade throughout the SoS. An SoS needs to be resilient to deal with the impact of complex internal and external environments. In this paper, we propose a resilience analysis method of an SoS based on a hierarchy structure. Firstly, we establish a hierarchy structure, which is ranked from high to low as capability level, activity level and system level. Then, Bayesian Networks (BNs) and Dynamic Bayesian Networks (DBNs) are used to analyze the resilience of the SoS. A resilience-based system importance metric is introduced, which is used in the budget allocation optimization problem during the development domain of an SoS. This paper proposes a mathematical programming model aimed at optimizing SoS resilience by optimally using budget to the subsystem. The application of the proposed approach is demonstrated using a case study: a Next Generation Air Transportation setting. The study results provide evidence that the proposed inter-dependency analysis based on Bayesian theory and the SoS resilience design approach can assist SoS system engineers in increasing expected SoS resilience during the development domain. [ABSTRACT FROM AUTHOR]

This paper presents the development of a Hybrid Model (HM) integrated with a Bayesian Network (BN) for comprehensive coastal vulnerability and risk assessment, with a focus on Konyaaltı Beach, Antalya, Turkey. The HM incorporates critical environmental parameters such as wind, waves, currents, and sediment transport to simulate conditions at vulnerable coastal areas and perform risk assessments for storm effects, flooding, and erosion. The model includes submodules for predicting coastal storms, quantifying sediment transport rates, assessing tsunami inundation severity, and categorizing storms based on beach typologies. The Adaptive Neuro-Fuzzy Inference System (ANFIS) is utilized for significant wave height predictions, enhancing the model's accuracy. The integration of hydrodynamic modeling, Bayesian networks, and ANFIS offers a robust framework for assessing coastal vulnerability and informing sustainable management practices. The study's results highlight the necessity for integrated risk management strategies, including adaptive infrastructure design, zoning and land use regulations, ecosystem-based management, and continuous monitoring and model refinement to enhance coastal resilience against dynamic environmental forces. This research provides valuable insights for mitigating the impacts of hazards on urban developments, contributing to the advancement of sustainable coastal management. [ABSTRACT FROM AUTHOR]

One of the challenges practitioners face when applying structure learning algorithms to their data involves determining a set of hyperparameters; otherwise, a set of hyperparameter defaults is assumed. The optimal hyperparameter configuration often depends on multiple factors, including the size and density of the usually unknown underlying true graph, the sample size of the input data, and the structure learning algorithm. We propose a novel hyperparameter tuning method, called the Out-of-sample Tuning for Structure Learning (OTSL), that employs out-of-sample and resampling strategies to estimate the optimal hyperparameter configuration for structure learning, given the input dataset and structure learning algorithm. Synthetic experiments show that employing OTSL to tune the hyperparameters of hybrid and score-based structure learning algorithms leads to improvements in graphical accuracy compared to the state-of-the-art. We also illustrate the applicability of this approach to real datasets from different disciplines. [ABSTRACT FROM AUTHOR]

To analyze the influencing factors of the emergency linkage of gas explosion accidents and their causal relationships, a method for analyzing the influencing factors of the emergency linkage of gas explosion accidents is proposed based on a hierarchical holographic model and Bayesian networks. Firstly, based on the hierarchical holographic model to determine the main influencing factors of the accident emergency linkage, we constructed the topological structure model of accident control, secondary disaster, and emergency rescue, and used the triangular fuzzy number to assess the a priori probability and conditional probability of the influencing factors. Next, the most likely factors affecting the accident emergency linkage are quickly identified by combining Bayesian diagnostic reasoning. Finally, sensitivity analysis is conducted to identify the key factors affecting the accident emergency linkage. The results show that the probability of normal operation of gas explosion accident emergency linkage is 78.1%, but when the coal mine environment changes, especially when multiple influencing factors occur simultaneously, the probability of normal operation of accident emergency linkage decreases significantly. Through causal analysis, the degree of influence on the operation of the accident emergency linkage in different situations can be deduced. Through diagnostic analysis, it can be seen that the emergency linkage operation is more sensitive to the two factors of the ventilation and smoke extraction system response and gas over limit, so it is necessary to pay attention to its important role in accident treatment. Meanwhile, the sensitivity analysis shows that the response of the ventilation and smoke extraction system, the accuracy of disaster sharing, the gas over limit, the technical level of the operators, and the team rescue experience are the key factors affecting the emergency response linkage in accidents. This study can provide theoretical guidance for the improvement of the emergency response linkage mechanism of coal mine gas explosion accidents as well as the decision-making of the accidents, minimize the losses of the accidents, and promote the sustainable development of the coal mining industry. [ABSTRACT FROM AUTHOR]

Optimizing CO2 injection for enhanced oil recovery (EOR) requires a precise estimation of the CO2-diffusivity coefficient in porous media. This study developed a predictive model for the molecular diffusivity coefficient of CO2 in bitumen and heavy crude oils using a Bayesian regularized artificial neural network. The unique contributions of the developed model compared to existing models include: First, a simple and accurate mathematical correlation between inputs and CO2-diffusivity has been presented, making it easy to use particularly for individuals with limited understanding of machine learning. Secondly, the time it takes the model to make a prediction (its inference latency) has been established and the model has been physically validated using trend analysis. Fourthly, independent datasets were used to test for the model generalizability.The model was evaluated using 260 data points from the literature, with 70 % for training and 30 % for testing. Key performance metrics were calculated, including root mean square error (RMSE = 0.03), and coefficient of determination (R2 = 0.996). Furthermore, an outlier detection analysis using the statistical leverage approach demonstrated that the data used was of high quality. A relevancy factor analysis revealed that pressure had the greatest impact on CO2 diffusivity, followed by temperature, while CO2 mass fraction had the least impact. The developed model operates within a temperature range of 295 K – 363 K and a pressure range of 1 MPa – 8MPa. Overall, the outcomes of this study contribute to the efficient prediction of CO2 diffusivity in heavy crude oil and bitumen.

Abstract While there are currently over 40 replicated genes with mapped risk alleles for Late Onset Alzheimer’s disease (LOAD), the Apolipoprotein E locus E4 haplotype is still the biggest driver of risk, with odds ratios for neuropathologically confirmed E44 carriers exceeding 30 (95% confidence interval 16.59–58.75). We sought to address whether the APOE E4 haplotype modifies expression globally through networks of expression to increase LOAD risk. We have used the Human Brainome data to build expression networks comparing APOE E4 carriers to non-carriers using scalable mixed-datatypes Bayesian network (BN) modeling. We have found that VGF had the greatest explanatory weight. High expression of VGF is a protective signal, even on the background of APOE E4 alleles. LOAD risk signals, considering an APOE background, include high levels of SPECC1L, HLA-DRA and RANBP3L. Our findings nominate several new transcripts, taking a combined approach to network building including known LOAD risk loci.

Bayesian networks (BNs) are probabilistic graphical models that leverage Bayes’ theorem to portray dependencies and cause-and-effect relationships between variables. These networks have gained prominence in the field of health sciences, particularly in diagnostic processes, by allowing the integration of medical knowledge into models and addressing uncertainty in a probabilistic manner. Objectives: This review aims to provide an exhaustive overview of the current state of Bayesian networks in disease diagnosis and prognosis. Additionally, it seeks to introduce readers to the fundamental methodology of BNs, emphasising their versatility and applicability across varied medical domains. Employing a meticulous search strategy with MeSH descriptors in diverse scientific databases, we identified 190 relevant references. These were subjected to a rigorous analysis, resulting in the retention of 60 papers for in-depth review. The robustness of our approach minimised the risk of selection bias. Results: The selected studies encompass a wide range of medical areas, providing insights into the statistical methodology, implementation feasibility, and predictive accuracy of BNs, as evidenced by an average area under the curve (AUC) exceeding 75%. The comprehensive analysis underscores the adaptability and efficacy of Bayesian networks in diverse clinical scenarios. The majority of the examined studies demonstrate the potential of BNs as reliable adjuncts to clinical decision-making. The findings of this review affirm the role of Bayesian networks as accessible and versatile artificial intelligence tools in healthcare. They offer a viable solution to address complex medical challenges, facilitating timely and informed decision-making under conditions of uncertainty. The extensive exploration of Bayesian networks presented in this review highlights their significance and growing impact in the realm of disease diagnosis and prognosis. It underscores the need for further research and development to optimise their capabilities and broaden their applicability in addressing diverse and intricate healthcare challenges.

Abstract Background The correlation between metals and hypertension, such as sodium, zinc, potassium, and magnesium, has been confirmed, while the relationship between aluminum and hypertension is not very clear. This study aimed to evaluate the correlation between plasma aluminum and hypertension in electrolytic aluminum workers by the Bayesian networks (BN). Methods In 2019, 476 male workers in an aluminum factory were investigated. The plasma aluminum concentration of workers was measured by inductively coupled plasma mass spectrometry. The influencing factors on the prevalence of hypertension were analyzed by the BN. Results The prevalence of hypertension was 23.9% in 476 male workers. The risk of hypertension from plasma aluminum in the Q2, Q3, and Q4 groups was 5.20 (1.90–14.25), 6.92 (2.51–19.08), and 7.33 (2.69–20.01), respectively, compared with that in the Q1 group. The risk of hypertension from the duration of exposure to aluminum of >10 years was 2.23 (1.09–4.57), compared without aluminum exposure. Area under the curve was 0.80 of plasma aluminum and the duration of exposure to aluminum was based on covariates, indicating that aluminum exposure had important predictive value in the prevalence of hypertension in the occupational population. The results of the study using the BN model showed that if the plasma aluminum of all participants was higher than Q4 (≥47.86 µg/L) and the participants were drinking, smoking, diabetes, central obesity, dyslipidemia, and aged >50 years, the proportion of hypertension was 71.2%. Conclusions The prevalence of hypertension increased significantly with the increase of plasma aluminum level.

Though network analysis has a long history in both natural and social sciences it has emerged as a new method in psychology in recent years. Unlike medical disorders, mental disorders are not observable in laboratory. However, we can identify them by the way of observable symptoms. According to the network perspective, a disorder occurs when an external event triggers a psychological symptom. Activated symptom also interacts with other symptoms and forms a pattern of symptoms. Network approach criticizes traditional categorical diagnostic approach and focuses on symptom organization. Probably, treating the most effective symptom will accelerate recovery process and provide more effective treatment. Network analysis can be used in both cross-sectional and longitudinal studies. Psychological networks provide opportunities to investigate direction of the relationship among symptoms, comorbidity, external triggers of psychological symptoms, effectiveness of treatment, comparison of symptom pattern according to sample characteristics. Despite the utility of psychological networks, accuracy of them has been questioned and certain methods to prove accuracy of networks proposed as response. Technological progress in recent years enabled network analysis to be more eligible in psychology. R Statistics software is very useful in network analysis which is totally free and open sourced and supported by many additional packages. This review article aims is to provide information about usage of network analysis in psychology, especially in clinical research. In the first part historical and theoretical background of network analysis was introduced and in the following parts structure, validity of psychological networks and R Statistics Software which is used for conducting network analysis were explained briefly.

Abstract Background The main goal of the paper is to define the level of energy efficiency in the economies of selected countries in the Balkan region that have opted for the EU Green Deal, a circular economy, and a transition to carbon neutrality. Energy efficiency, as a determinant of carbon neutrality, was selected as an indicator for analysis because it records particularly unfavorable indicators in the region under observation. The research was carried out on a sample of seven Balkan countries and their surrounding areas. An initial qualitative analysis was followed by a quantitative analysis based on a combination of statistical methods and soft computing. Six indicators were selected for the analysis covering a period of 30 years (1990–2020). Results A significant obstacle to the green transition and the region’s transition to a circular economy and carbon neutrality is energy efficiency and energy related pollution—the reliance of most countries on coal-fired thermal power plants for electricity generation. The research results showed the following: (a) the degree of economic development and membership in the European Union are not significantly related to the level of energy efficiency; (b) most of the sampled countries are in the initial stages of introducing activities to achieve carbon neutrality; and (c) only Slovenia has documented consistent indicators and evident advancements in its efforts to achieve carbon neutrality. Based on the research findings, proposals for improvements were made in the direction of policymaking and in a methodological sense. Conclusions The implementation of circularity and carbon neutrality as a long-term goal of the European Union is not necessarily related to the level of economic development, nor can its trajectory be exclusively ascertained by means of data processing and monitoring. A more precise understanding of a carbon-neutral future can be achieved through the incorporation of qualitative data to a greater extent, a realistic evaluation of historical facts and their repercussions, as well as projections of the effects that reality and global developments after 2022 will have on each country.

Researchers studying the evolution of viral pathogens and other organisms increasingly encounter and use large and complex data sets from multiple different sources. Statistical research in Bayesian phylogenetics has risen to this challenge. Researchers use phylogenetics not only to reconstruct the evolutionary history of a group of organisms, but also to understand the processes that guide its evolution and spread through space and time. To this end, it is now the norm to integrate numerous sources of data. For example, epidemiologists studying the spread of a virus through a region incorporate data including genetic sequences (e.g. DNA), time, location (both continuous and discrete) and environmental covariates (e.g. social connectivity between regions) into a coherent statistical model. Evolutionary biologists routinely do the same with genetic sequences, location, time, fossil and modern phenotypes, and ecological covariates. These complex, hierarchical models readily accommodate both discrete and continuous data and have enormous combined discrete/continuous parameter spaces including, at a minimum, phylogenetic tree topologies and branch lengths. The increased size and complexity of these statistical models have spurred advances in computational methods to make them tractable. We discuss both the modeling and computational advances below, as well as unsolved problems and areas of active research.

For contaminated sites, conceptual site models (CSMs) guide the assessment and management of risks, including remediation strategies. Recent research has expanded diagrammatic CSMs with structural causal modeling to develop what are nominally called conceptual Bayesian networks (CBNs) for environmental risk assessment. These CBNs may also be useful for problems of controlling and preventing offsite contaminant migration, especially for sites containing dense nonaqueous phase liquids (DNAPLs). In particular, the CBNs provide greater clarity on the causal relationships between source term, onsite and offsite migration, and remediation effectiveness characterization for contaminated DNAPL sites compared to traditional CSMs. These ideas are demonstrated by the inclusion of modifying variables, causal pathway analysis, and interventions in CBNs. Additionally, several new extensions of the CBN concept are explored including the representation of measurement variables as lines of evidence and alignment with conventional pictorial CSMs for groundwater modeling. Taken as a whole, the CBNs provide a powerful and adaptable knowledge representation tool for remediating subsurface systems contaminated by DNAPL. [ABSTRACT FROM AUTHOR]

In recent years, mixed Bayesian networks have received increasing attention across various fields for probabilistic reasoning. Though many studies have been devoted to propagation computation on strong junction trees for mixed Bayesian networks, few have addressed the construction of appropriate strong junction trees. In this work, we establish a connection between the minimal strong triangulation for marked graphs and the minimal triangulation for star graphs. We further propose a minimal strong triangulation method for the moral graph of mixed Bayesian networks and develop a polynomial-time algorithm to derive a strong junction tree from this minimal strong triangulation. Moreover, we also focus on the propagation computation of all posteriors on this derived strong junction tree. We conducted multiple numerical experiments to evaluate the performance of our proposed method, demonstrating significant improvements in computational efficiency compared to existing approaches. Experimental results indicate that our minimal strong triangulation approach provides a robust framework for efficient probabilistic inference in mixed Bayesian networks. [ABSTRACT FROM AUTHOR]

Resilience theory is introduced in the present study to model the capability of the Arctic shipping system to defend against various risks. For this purpose, a novel methodology integrating a Bayesian network (BN), system dynamics (SD), and dissipative structure is proposed to quantitatively describe the dynamic resilience of the Arctic shipping system. The evolution of the system resilience with time is simulated, based on which the graphical results and discussion are devoted to revealing the dynamic evolution mechanism by integrating the three aspects of triggering conditions, priorities, and driving forces. Overall, the proposed resilience evolution model is suitable for complex systems with uncertainty and nonlinearity, and the results show that the proposed methodology can contribute to formulating future strategies and decisions to guarantee the safety of Arctic shipping. [ABSTRACT FROM AUTHOR]

The increasing prevalence of species that are detrimental to biodiversity is a major concern, particularly for managers of national parks. To develop effective programmes for controlling weeds, it is essential to have a thorough understanding of the extent and severity of infestations, as well as the contributing factors such as temperature, rainfall, and disturbance. Predicting these factors on a regional scale requires models that can incorporate a wide range of variables in a quantifiable manner, while also assisting with on-ground operations. In this study, we present two Bayesian Network models specifically designed for six significant weed species found along the southern coast of Australia. Our models are based on empirical data collected during a coastal weed survey conducted in 2015 and repeated in 2016. We applied these models to the coastal national parks in the isolated and pristine East Gippsland region. Importantly, the prediction models were developed at two different spatial scales that directly corresponded to the scale of the observations. Our findings indicate that coastal habitats, with their vulnerable environments and prevalence of open dune systems, are particularly susceptible to weed infestations. Moreover, adjacent regions also have the potential for colonization if these infestations are not effectively controlled. Climate-related factors play a role in moderating the potential for colonization, which is a significant concern for weed control efforts in the context of global climate change. [ABSTRACT FROM AUTHOR]

Causal Bayesian Networks (CBNs) provide an important tool for reasoning under uncertainty with potential application to many complex causal systems. Structure learning algorithms that can tell us something about the causal structure of these systems are becoming increasingly important. In the literature, the validity of these algorithms is often tested for sensitivity over varying sample sizes, hyper-parameters, and occasionally objective functions, but the effect of the order in which the variables are read from data is rarely quantified. We show that many commonly-used algorithms, both established and state-of-the-art, are more sensitive to variable ordering than these other factors when learning CBNs from discrete variables. This effect is strongest in hill-climbing and its variants where we explain how it arises, but extends to hybrid, and to a lesser-extent, constraint-based algorithms. Because the variable ordering is arbitrary, any significant effect it has on learnt graph accuracy is concerning, and raises questions about the validity of both many older and more recent results produced by these algorithms in practical applications and their rankings in performance evaluations. [ABSTRACT FROM AUTHOR]

Context. The problem of human factor management in the process of making relevant decisions in the management of critical infrastructure facilities is currently very important and complex. This issue is becoming increasingly significant due to the dynamic and unpredictable nature of the environment in which these facilities operate. Effective management of CIF requires the development of new models and methods that are based on adaptive management principles. These models and methods must take into account the personal emotional and cognitive capabilities of the decision maker, who is often operating under the influence of destabilizing uncertain factors. The challenge is further compounded by the need to integrate these adaptive methods into existing human-machine systems, ensuring that they can respond in real-time to the rapidly changing conditions that can affect the decision-making process. The complexity and importance of this problem necessitate a multifaceted approach that combines probabilistic methods, intellectual technologies, and information-cognitive technologies. These technologies must be capable of providing real-time adaptation and assessment of the DM’s emotional and cognitive state, which is critical for making relevant and timely decisions. The current unresolved problems in the field of creating adaptive information technologies for decision support in the management of CIF highlight the urgent need for a promising approach that can address these issues effectively and efficiently. Objective. The objective is to propose a a comprehensive method for evaluating the process of relevant decision-making, which depends on the functional stability of critical infrastructure facilities and the adaptation of factors related to the emotional-cognitive state of the decision maker. This method aims to provide a systematic approach to understanding how various factors, including the psycho-functional state of the DM, influence the decision-making process. Additionally, the objective includes the development of adaptive information and intellectual technologies that can support real-time evaluation and adjustment of the DM’s emotional and cognitive states. This approach seeks to ensure that decisions are made efficiently and effectively, even under the influence of destabilizing uncertain factors. By addressing these aspects, the method aims to enhance the overall reliability and resilience of the CIF management processes. Furthermore, the objective encompasses the integration of Bayesian networks and a comprehensive knowledge base to facilitate the decision support system in providing timely and accurate information for decision-making. Method. To implement this method, probabilistic methods, intellectual and information-cognitive technologies were used to provide acceptable adaptation and evaluation of the relevant decision-making process in real-time. Results. The proposed method, based on intellectual and information-cognitive technology, allows for real-time assessment and adaptation of the emotional and cognitive state of the decision maker during the process of making relevant decisions. The implementation of probabilistic methods and Bayesian networks has enabled the development of a robust decision support system that effectively integrates adaptive management principles. This system ensures that the decision-making process remains stable and reliable, even in the presence of destabilizing uncertain factors. The real-time capabilities of the system allow for prompt adjustments to the psycho-functional state of the DM, which is critical for maintaining the functional stability of critical infrastructure facilities. The results demonstrate that the use of intellectual technologies and a comprehensive knowledge base significantly enhances the DM’s ability to make informed decisions. Experiments have shown that this method improves the overall efficiency and effectiveness of CIF management, providing a promising approach for future applications in adaptive decision support processes. The results obtained from these experiments validate the potential of the proposed method to revolutionize the management of CIF by ensuring that decisions are both timely and appropriate, thereby contributing to the resilience and reliability of these essential facilities.. Conclusions. The results of the experiments allow us to recommend the use of the proposed method of rapid assessment and adaptation of the emotional and cognitive state of the decision maker for the process of making relevant decisions in real-time. The integration of intellectual and information-cognitive technologies into the decision support system has proven to be effective in enhancing the stability and reliability of the decision-making process in the management of critical infrastructure facilities. The realtime capabilities of the system facilitate prompt adjustments to the psycho-functional state of the DM, ensuring that decisions are made efficiently and effectively, even under the influence of destabilizing uncertain factors. The experimental results demonstrate that the proposed method significantly improves the overall efficiency of CIF management by providing a robust framework for adaptive decision support. The results obtained can be used in the development of adaptive DSS in the management of CIF, offering a promising approach for future applications. This method not only enhances the decision-making capabilities of DMs but also contributes to the resilience and reliability of CIF, ensuring their functional stability in dynamic and uncertain environments. [ABSTRACT FROM AUTHOR]

Gradient boosting machines harnesses the inherent capabilities of decision trees and meticulously corrects their errors in a sequential fashion, culminating in remarkably precise predictions. Word2Vec, a prominent word embedding technique, occupies a pivotal role in natural language processing (NLP) tasks. Its proficiency lies in capturing intricate semantic relationships among words, thereby facilitating applications such as sentiment analysis, document classification, and machine translation to discern subtle nuances present in textual data. Bayesian networks introduce probabilistic modeling capabilities, predominantly in contexts marked by uncertainty. Their versatile applications encompass risk assessment, fault diagnosis, and recommendation systems. Gated recurrent units (GRU), a variant of recurrent neural networks, emerges as a formidable asset in modeling sequential data. Both training and testing are crucial to the success of an intrusion detection system (IDS). During the training phase, several models are created, each of which can recognize typical from anomalous patterns within a given dataset. To acquire passwords and credit card details, "phishing" usually entails impersonating a trusted company. Predictions of student performance on academic tasks are improved by hyper parameter optimization of the gradient boosting regression tree using the grid search approach. [ABSTRACT FROM AUTHOR]

In this article, the authors focus on the introduction of a hybrid method for risk-based fault detection (FD) using dynamic principal component analysis (DPCA) and failure method and effect analysis (FMEA) based Bayesian networks (BNs). The FD problem has garnered great interest in industrial application, yet methods for integrating process risk into the detection procedure are still scarce. It is, however, critical to assess the risk each possible process fault holds to differentiate between non-safety-critical and safety-critical abnormalities and thus minimize alarm rates. The proposed method utilizes a BN established through FMEA analysis of the supervised process and the results of dynamical principal component analysis to estimate a modified risk priority number (RPN) of different process states. The RPN is used parallel to the FD procedure, incorporating the results of both to differentiate between process abnormalities and highlight critical issues. The method is showcased using an industrial benchmark problem as well as the model of a reactor utilized in the emerging liquid organic hydrogen carrier (LOHC) technology. [ABSTRACT FROM AUTHOR]

In order to realize the intelligent inference of structural vulnerability, Bayesian networks(BNs)have been adopted for reconstructing the analysis system of a truss structure. Firstly, the external load combination, the truss system and its members are defined as the top parent nodes, the middle nodes and the bottom child nodes of the network, respectively. These nodes are connected by some directed edges representing the causality between them. Therefore, the BN topology of the truss structure is defined. Secondly, serious damage is suggested to replace the commonly-used assumption of conceptual removal in vulnerability analysis. Based on the uncertainty distributions of the parameters of members and external loads, the samples are randomly drawn from the probability distributions for learning the conditional probability tables between the different nodes, thereby realizing the BN establishment. Thirdly, the observed state of a specific member is used as the evidence into the established BN for synchronously inferring the state probabilities of the other members, based on which the member importance coefficient is calculated. The sum of all the importance coefficients is further defined as the system vulnerability index. Finally, a member vulnerability index is proposed to predict the most probable failure path of the truss system. The numerical and experimental examples have demonstrated that the proposed method can effectively evaluate the importance of each member within a truss system. The inferred failure path accords well with the experimental observation. The estimated value of the system vulnerability index of the experimental truss is far less than the number of members, and thus, it indicates the low possibility of progressive collapse of the system due to the damaged member. [ABSTRACT FROM AUTHOR]

A strategy refers to the rules that the agent chooses the available actions to achieve goals. Adopting reasonable strategies is challenging but crucial for an intelligent agent with limited resources working in hazardous, unstructured, and dynamic environments to improve the system's utility, decrease the overall cost, and increase mission success probability. This article proposes a novel hierarchical strategy decomposition approach based on Bayesian chaining to separate an intricate policy into several simple sub-policies and organize their relationships as Bayesian strategy networks (BSN). We integrate this approach into the state-of-the-art DRL method—soft actor-critic (SAC), and build the corresponding Bayesian soft actor-critic (BSAC) model by organizing several sub-policies as a joint policy. Our method achieves the state-of-the-art performance on the standard continuous control benchmarks in the OpenAI Gym environment. The results demonstrate that the promising potential of the BSAC method significantly improves training efficiency. Furthermore, we extend the topic to the Multi-Agent systems (MAS), discussing the potential research fields and directions. [ABSTRACT FROM AUTHOR]

Based on train operation data, Bayesian networks (BN) are used to model the cascading effects of traffic control actions and their influences (such as changes in train delays) for two and three consecutive trains. Influence factors are first determined to describe the interactions between train delays and control actions, i.e. the delay changes due to control actions over time and space. Considering the interdependence of these factors, their causal relationships are obtained, and the BN model's connection paradigm is determined based on these causal ties. The BN structures are then proposed by combining domain knowledge and a data-driven method. The proposed models are tested on the train operation data from the Chinese high-speed railways. The results show that the proposed method exhibits a good fit for the train operation data and outperforms other conventional train operation models in terms of various evaluation metrics. Besides, the strength of train control actions' cascading effects is investigated. It shows that section train control actions are stronger than those in stations, and both are considerably correlated with train delays and recovery times in sections and stations. Finally, a macroscopic model for control actions between several trains (more than three) is obtained based on the learned paradigm of the presented models and the cascading effects between two and three trains, demonstrating the model's extensibility. The proposed models are intended to support traffic controllers with the estimation of future train delay change patterns, the expected control actions, and the cascading effects of control actions, and they are imperative for aiding the decision-making of controllers to manage high-speed railway traffic. [ABSTRACT FROM AUTHOR]

Copyright of Automatisierungstechnik is the property of De Gruyter and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Background: Business continuity entails the potential negative consequences of uncertainty on a firm's ability to achieve strategic objectives. The COVID-19 pandemic significantly impacted business continuity due to lockdowns, travel restrictions, and social distancing measures. Consequently, firms adopted specific supply chain (SC) practices to effectively navigate this global crisis. Methods: This research adopted a stochastic approach based on Bayesian Networks to evaluate the implications of business continuity on firms' decisions to embrace SC practices, focusing on omnichannel strategies, SC coordination, and technologies such as artificial intelligence systems, big data and machine learning, and mobile applications. Results: Our findings revealed that firms facing disruption in a single performance area can apply specific strategies to maintain resilience. However, multiple areas of underperformance necessitate a varied approach. Conclusions: According to our empirical analysis, omnichannel strategies are critical when disruptions simultaneously impact quality, inventory, sales, and ROI, particularly during major disruptions such as the COVID-19 pandemic. AI and big data become vital when multiple risks coalesce, enhancing areas such as customer service and supply chain visibility. Moreover, supply chain coordination and mobile app adoption are effective against individual performance risks, proving crucial in mitigating disruption impacts across various business aspects. These findings help policy-makers and business owners to have a better understanding of how business continuity based on performance resistance to disruptions pushes companies to adopt different practices including new technologies and supply chain coordination. Accordingly, they can use the outputs of this study to devise strategies for improving resilience considering their supply chain vulnerabilities. [ABSTRACT FROM AUTHOR]

Bayesian networks (BNs) are probabilistic graphical models that leverage Bayes' theorem to portray dependencies and cause-and-effect relationships between variables. These networks have gained prominence in the field of health sciences, particularly in diagnostic processes, by allowing the integration of medical knowledge into models and addressing uncertainty in a probabilistic manner. Objectives: This review aims to provide an exhaustive overview of the current state of Bayesian networks in disease diagnosis and prognosis. Additionally, it seeks to introduce readers to the fundamental methodology of BNs, emphasising their versatility and applicability across varied medical domains. Employing a meticulous search strategy with MeSH descriptors in diverse scientific databases, we identified 190 relevant references. These were subjected to a rigorous analysis, resulting in the retention of 60 papers for in-depth review. The robustness of our approach minimised the risk of selection bias. Results: The selected studies encompass a wide range of medical areas, providing insights into the statistical methodology, implementation feasibility, and predictive accuracy of BNs, as evidenced by an average area under the curve (AUC) exceeding 75%. The comprehensive analysis underscores the adaptability and efficacy of Bayesian networks in diverse clinical scenarios. The majority of the examined studies demonstrate the potential of BNs as reliable adjuncts to clinical decision-making. The findings of this review affirm the role of Bayesian networks as accessible and versatile artificial intelligence tools in healthcare. They offer a viable solution to address complex medical challenges, facilitating timely and informed decision-making under conditions of uncertainty. The extensive exploration of Bayesian networks presented in this review highlights their significance and growing impact in the realm of disease diagnosis and prognosis. It underscores the need for further research and development to optimise their capabilities and broaden their applicability in addressing diverse and intricate healthcare challenges. [ABSTRACT FROM AUTHOR]

This study aims to address quality issues in the production of prefabricated steel structural components for buildings by investigating challenges in quality risk assessment. It identifies key factors contributing to quality problems and establishes an evaluation index system. Traditional methods encounter limitations in handling uncertainty and conducting quantitative analysis. Therefore, the fuzzy Bayesian network (FBN) theory is utilized to perform a probabilistic analysis of quality risks during the production phase. This research achieves a more accurate and dynamic risk assessment by integrating the strengths of fuzzy logic and Bayesian networks (BNs) and by utilizing expert knowledge, the similarity aggregation method (SAM), and the noisy-OR gate model. The study reveals that factors such as the "low professional level of designers", "poor production refinement", and "poor storage conditions for finished products" have a significant impact on quality risks. This study offers a scientific risk assessment tool designed to address the quality control challenges commonly experienced in the manufacturing of steel structural components. Identifying the critical risk factors that influence quality empowers actual production enterprises to develop risk management strategies and improvement measures in a more focused manner, thereby facilitating more effective resource allocation and risk prevention and control. Consequently, this approach has a significant impact on enhancing the overall production level and quality within the industry. [ABSTRACT FROM AUTHOR]

The construction of buildings invariably involves time and costs, and disruptions impact ongoing construction projects. Crisis situations in management strategies, structural confusion, and financial miscalculations often arise due to misguided decision-making. This article proposes a method that combines the learning of Bayesian Networks and heuristic techniques to optimize decision-making processes in construction scheduling. As an innovative approach in order to enhance construction management, the functioning of biological, molecular, and physical objects and nervous systems is considered, applying bionic features to mimic their efficiency and precision, thereby optimizing construction processes and improving coordination and decision-making. Bayesian Networks are used for probabilistic analysis, and heuristic methods guide quick decision-making. The results demonstrate the effectiveness of Bayesian Networks and heuristic methods in data analysis and decision-making in construction engineering. The developed algorithm can be successfully applied to both erecting and planning construction projects. [ABSTRACT FROM AUTHOR]

Bayesian networks are commonly used for learning with uncertainty and incorporating expert knowledge. However, they are hard to interpret, especially when the network structure is complex. Methods used to explain Bayesian networks operate under certain assumptions about what constitutes the best explanation, without actually verifying these assumptions. One such common assumption is that a shorter length of the causal chain of one variable to another enhances its explanatory strength. Counterfactual explanations gained popularity in artificial intelligence over the last years. It is well-known that it is possible to generate counterfactuals from causal Bayesian networks, but there is no indication which of them are useful for explanatory purposes. In this paper, we examine how to apply findings from psychology to search for counterfactuals that are perceived as more useful explanations for the end user. For this purpose, we have conducted a questionnaire to test whether counterfactuals that change an actionable cause are considered more useful than counterfactuals that change a direct cause. The results of the questionnaire indicate that actionable counterfactuals are preferred regardless of being the direct cause or having a longer causal chain. [ABSTRACT FROM AUTHOR]

While preeclampsia is the leading cause of maternal death in Guayas province (Ecuador), its causes have not yet been studied in depth. The objective of this research is to build a Bayesian network classifier to diagnose cases of preeclampsia while facilitating the understanding of the causes that generate this disease. Data for the years 2017 through 2023 were gathered retrospectively from medical histories of patients treated at "IESS Los Ceibos" hospital in Guayaquil, Ecuador. Naïve Bayes (NB), The Chow–Liu Tree-Augmented Naïve Bayes ( TAN c l ), and Semi Naïve Bayes (FSSJ) algorithms have been considered for building explainable classification models. A proposed Non-Redundant Feature Selection approach (NoReFS) is proposed to perform the feature selection task. The model trained with the TAN c l and NoReFS was the best of them, with an accuracy close to 90%. According to the best model, patients whose age is above 35 years, have a severe vaginal infection, live in a rural area, use tobacco, have a family history of diabetes, and have had a personal history of hypertension are those with a high risk of developing preeclampsia. [ABSTRACT FROM AUTHOR]

Human lifestyles vary enormously over time and space and so understanding the origins of this diversity has always been a central focus of anthropology. A major source of this cultural variation is the variation in institutional complexity: the cultural packages of rules, norms, ontologies and expectations passed down through societies across generations. In this article, we study the emergence of institutions in small-scale societies. There are two primary schools of thought. The first is that institutions emerge top-down as rules are imposed by elites on their societies in order to gain asymmetrical access to power, resources and influence over others through coercion. The second is that institutions emerge bottom-up to facilitate interactions within populations as they seek collective solutions to adaptive problems. Here, we use Bayesian networks to infer the causal structure of institutional complexity in 172 small-scale societies across ethnohistoric western North America reflecting the wide diversity of indigenous lifestyles across this vast region immediately prior to European colonization. Our results suggest that institutional complexity emerges from underlying socioecological complexity because institutions are solutions to coordination problems in more complex environments where human–environment interactions require increased management. [ABSTRACT FROM AUTHOR]

• A fusion-driven Bayesian network fault diagnosis method is proposed. • Broad learning system-based benchmark model is developed to learn prior structure. • Fuzzy logic systems-based knowledge evaluation method is developed for structuring optimization. • Fusion field observation information to improve the accuracy and reliability of fault diagnosis. • The method was experimented on a chiller. Timely diagnosis and maintenance of faults in chiller units is beneficial for reducing energy consumption in buildings. In order to improve the diagnostic accuracy and provide reasonable explanations for fault occurrences of the heating, ventilation, and air conditioning (HVAC) systems in buildings, a fusion-driven Bayesian network fault diagnosis method is proposed based on broad learning system and fuzzy expert knowledge. Firstly, from the data-driven perspective, an efficient broad learning system is developed as benchmarking model to build the prior Bayesian network structure. Secondly, from the knowledge-driven perspective, the fuzzy logic system is used to fuzz expert knowledge, which then used to optimize the Bayesian network. Finally, the information obtained by experts on-site is incorporated into the optimized Bayesian network as new evidence nodes to determine the Bayesian network. The parameters of the Bayesian network are learned through Noisy-MAX processing of the conditional probability table. The performance of the proposed method is evaluated based on the fault diagnosis of a chiller system. The results demonstrate that the Bayesian network established through this method combines the advantages of data-driven and knowledge-driven approaches. It not only performs well in terms of diagnostic accuracy, with accuracy rates above 98 % for six typical faults, but also provides effective explanations for the underlying mechanisms of the faults through causal relationship diagrams. [ABSTRACT FROM AUTHOR]

Constructing Bayesian networks (BN) for practical applications presents significant challenges, especially in domains with limited empirical data available. In such situations, field experts are often consulted to estimate the model's parameters, for instance, rank correlations in Gaussian copula-based Bayesian networks (GCBN). Because there is no consensus on a 'best' approach for eliciting these correlations, this paper proposes a framework that uses probabilities of concordance for assessing dependence, and the dependence calibration score to aggregate experts' judgments. To demonstrate the relevance of our approach, the latter is implemented to populate a GCBN intended to estimate the condition of air handling units' components—a key challenge in building asset management. While the elicitation of concordance probabilities was well received by the questionnaire respondents, the analysis of the results reveals notable disparities in the experts' ability to quantify uncertainty. Moreover, the application of the dependence calibration aggregation method was hindered by the absence of relevant seed variables, thus failing to evaluate the participants' field expertise. All in all, while the authors do not recommend to use the current model in practice, this study suggests that concordance probabilities should be further explored as an alternative approach for the elicitation of dependence. [ABSTRACT FROM AUTHOR]

Background: The main goal of the paper is to define the level of energy efficiency in the economies of selected countries in the Balkan region that have opted for the EU Green Deal, a circular economy, and a transition to carbon neutrality. Energy efficiency, as a determinant of carbon neutrality, was selected as an indicator for analysis because it records particularly unfavorable indicators in the region under observation. The research was carried out on a sample of seven Balkan countries and their surrounding areas. An initial qualitative analysis was followed by a quantitative analysis based on a combination of statistical methods and soft computing. Six indicators were selected for the analysis covering a period of 30 years (1990–2020). Results: A significant obstacle to the green transition and the region's transition to a circular economy and carbon neutrality is energy efficiency and energy related pollution—the reliance of most countries on coal-fired thermal power plants for electricity generation. The research results showed the following: (a) the degree of economic development and membership in the European Union are not significantly related to the level of energy efficiency; (b) most of the sampled countries are in the initial stages of introducing activities to achieve carbon neutrality; and (c) only Slovenia has documented consistent indicators and evident advancements in its efforts to achieve carbon neutrality. Based on the research findings, proposals for improvements were made in the direction of policymaking and in a methodological sense. Conclusions: The implementation of circularity and carbon neutrality as a long-term goal of the European Union is not necessarily related to the level of economic development, nor can its trajectory be exclusively ascertained by means of data processing and monitoring. A more precise understanding of a carbon-neutral future can be achieved through the incorporation of qualitative data to a greater extent, a realistic evaluation of historical facts and their repercussions, as well as projections of the effects that reality and global developments after 2022 will have on each country. [ABSTRACT FROM AUTHOR]

Artificial Intelligence (AI) and Machine Learning (ML) approaches that could learn from large data sources have been identified as useful tools to support clinicians in their decisional process; AI and ML implementations have had a rapid acceleration during the recent COVID-19 pandemic. However, many ML classifiers are "black box" to the final user, since their underlying reasoning process is often obscure. Additionally, the performance of such models suffers from poor generalization ability in the presence of dataset shifts. Here, we present a comparison between an explainable-by-design ("white box") model (Bayesian Network (BN)) versus a black box model (Random Forest), both studied with the aim of supporting clinicians of Policlinico San Matteo University Hospital in Pavia (Italy) during the triage of COVID-19 patients. Our aim is to evaluate whether the BN predictive performances are comparable with those of a widely used but less explainable ML model such as Random Forest and to test the generalization ability of the ML models across different waves of the pandemic. [ABSTRACT FROM AUTHOR]

Piracy and armed robbery continue to pose significant security threats to the shipping industry. This paper presents a real-time threat assessment and crisis classification tool for piracy or armed robbery incidents. The tool is part of a crisis classification module that addresses various categories of security threats. This module is currently being developed as part of the EU-funded research project ISOLA, which aims to introduce an intelligent security superintendence ecosystem. The ecosystem is designed to complement the existing ship security processes and measures applied onboard passenger ships. The tool operates by providing real-time threat classification and subsequent warnings by analysing data collected from the ship's legacy systems and installed sensors with the utilisation of Bayesian probabilistic techniques, particularly Bayesian Networks. The BN model developed for this purpose is thoroughly examined, and its validation is presented through indicative case studies involving piracy and armed robbery. The main objective is to improve situational awareness, enhance vigilance and early threat detection, and support the decision-making process for the Master and crew, especially under time-sensitive circumstances and stressful conditions. [ABSTRACT FROM AUTHOR]

Bayesian networks faithfully represent the symmetric conditional independences existing between the components of a random vector. Staged trees are an extension of Bayesian networks for categorical random vectors whose graph represents non-symmetric conditional independences via vertex coloring. However, since they are based on a tree representation of the sample space, the underlying graph becomes cluttered and difficult to visualize as the number of variables increases. Here, we introduce the first structural learning algorithms for the class of simple staged trees, entertaining a compact coalescence of the underlying tree from which non-symmetric independences can be easily read. We show that data-learned simple staged trees often outperform Bayesian networks in model fit and illustrate how the coalesced graph is used to identify non-symmetric conditional independences. [ABSTRACT FROM AUTHOR]

The Bayesian network provides a useful way to deal with uncertain information, which helps researchers to better understand the human cognitive process. The foundation of the Bayesian network focuses on identifying the qualitative relations between variables, which is also called structure learning. Local search in the ordering space is an effective method for learning the structure of large-scale Bayesian networks. However, the existing algorithms tend to the local optimum and stop searching for superior solutions. To tackle the problem, random perturbations are applied to the local optimum without specific strategies, resulting in many meaningless restarts that sacrifice much time but still fail to improve the results. As an extension of the local search, simulated annealing stochastically searches the solution spaces and selects relatively poor solutions with a certain probability. This paper proposes a method based on simulated annealing to learn Bayesian network structure in the ordering space, which expands the search scope by probabilistically accepting poorer solutions. Moreover, we improve simulated annealing by adding a memory module and modifying the termination condition. The memory module records the optimal solution before accepting a worse solution, which avoids losing the possible global optimal solution. The new termination condition is related to the quality of the search results, which reduces many redundant searches. Besides, we design a new restart strategy based on ensemble learning. When the search traps in the local optimum, a new ordering is obtained to restart the search by perturbing the current ordering with constraints. The constraints are generated by the results of ensemble learning on multiple structures, which help the algorithm approach the global optimum solution. Experimental results show that our proposed methods improve the accuracy and efficiency in learning the optimal structure over the benchmarks compared to the state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

Deep Learning (DL) stands out as a leading model for processing high-dimensional data, where the nonlinear transformation of hidden layers effectively extracts features. However, these unexplainable features make DL a low interpretability model. Conversely, Bayesian network (BN) is transparent and highly interpretable, and it can be helpful for interpreting DL. To improve the interpretability of DL from the perspective of feature cognition, we propose the feature analysis network (FAN), a DL structure fused with BN. FAN retains the DL feature extraction capability and applies BN as the output layer to learn the relationships between the features and the outputs. These relationships can be probabilistically represented by the structure and parameters of the BN, intuitively. In a further study, a correlation clustering-based feature analysis network (cc-FAN) is proposed to detect the correlations among inputs and to preserve this information to explain the features' physical meaning to a certain extent. To quantitatively evaluate the interpretability of the model, we design the network simplification and interpretability indicators separately. Experiments on eight datasets show that FAN has better interpretability than that of the other models with basically unchanged model accuracy and similar model complexities. On the radar effect mechanism dataset, from the feature structure-based relevance interpretability indicator, FAN is up to 4.8 times better than that of the other models, and cc-FAN is up to 21.5 times better than that of the other models. FAN and cc-FAN enhance the interpretability of the DL model structure from the aspects of features; moreover, based on the input correlations, cc-FAN can help us to better understand the physical meaning of features. [ABSTRACT FROM AUTHOR]