Bayesian Network Research Articles

Adverse Event Assessment of Upadacitinib: A Pharmacovigilance Study Based on the FAERS Database.

Upadacitinib, a Janus kinase (JAK) inhibitor, has been approved by the FDA to treat various autoimmune conditions. This study assessed its adverse events by analyzing reports from the FDA Adverse Event Reporting System (FAERS). FAERS data from Q3 2019 to Q4 2023 were extracted, and disproportionality analyses were conducted using four statistical measures, reporting odds ratio, proportionate reporting ratio, Bayesian confidence propagation neural network, and empirical Bayesian geometric mean. A total of 6 879 398 adverse event reports were collected, with 37 700 reports identifying upadacitinib as the "primary suspected." These reports involved 24 system organ classes and 246 preferred terms that met the criteria across all four algorithms. The distribution of adverse events was assessed separately for female and male patients. Further analysis of the top 25 preferred terms revealed that, although the system organ classes were similar between sexes, the specific adverse events differed. The adverse events were analyzed by gender, showing musculoskeletal and skin disorders were prevalent and severe in male patients, while musculoskeletal issues, infections, and abnormal laboratory tests were common in female patients. Unexpected events like trigger finger, biliary sepsis, and serious events such as oral neoplasm were also identified. This study provides real-world evidence for the safety evaluation of upadacitinib and underscores the need to monitor sex-specific adverse events. Future prospective studies are necessary to confirm these pharmacovigilance findings.

ANN uncertainty estimates in assessing fatty liver content from ultrasound data

Background and objectiveThis article uses three different probabilistic convolutional architectures applied to ultrasound image analysis for grading Fatty Liver Content (FLC) in Metabolic Dysfunction Associated Steatotic Liver Disease (MASLD) patients. Steatosis is a new silent epidemic and its accurate measurement is an impelling clinical need, not only for hepatologists, but also for experts in metabolic and cardiovascular diseases. This paper aims to provide a robust comparison between different uncertainty quantification strategies to identify advantages and drawbacks in a real clinical setting. MethodsWe used a classical Convolutional Neural Network, a Monte Carlo Dropout, and a Bayesian Convolutional Neural Network with the goal of not only comparing the goodness of the predictions, but also to have access to an evaluation of the uncertainty associated with the outputs. ResultsWe found that even if the prediction based on a single ultrasound view is reliable (relative RMSE [5.93%-12.04%]), networks based on two ultrasound views outperform them (relative RMSE [5.35%-5.87%]). In addition, the results show that the introduction of a “not confident” category contributes to increase the percentage of correctly predicted cases and to decrease the percentage of mispredicted cases, especially for semi-intrusive methods. ConclusionsThe possibility of having access to information about the confidence with which the network produces its outputs is a great advantage, both from the point of view of physicians who want to use neural networks as computer-aided diagnosis, and for developers who want to limit overfitting and obtain information about dataset problems in terms of out-of-distribution detection.

Redshift Prediction with Images for Cosmology Using a Bayesian Convolutional Neural Network with Conformal Predictions

In the emerging era of big data astrophysics, large-scale extragalactic surveys will soon provide high-quality imaging for billions of celestial objects to answer major questions in astrophysics such as the nature of dark matter and dark energy. Precision cosmology with surveys requires accurate photometric redshift (photo-z) estimation with well-constrained uncertainties as inputs for weak lensing models to measure cosmological parameters. Machine learning methods have shown promise in optimizing the information gained from galaxy images in photo-z estimation; however, many of these methods are limited in their ability to estimate accurate uncertainties. In this work, we present one of the first applications of Bayesian convolutional neural networks (BCNNs) for photo-z estimation and uncertainties. In addition, we use conformal mapping to calibrate the photo-z uncertainties to achieve good statistical coverage. We use the public GalaxiesML data set of ∼300k galaxies from the Hyper Suprime-Cam survey containing five-band photometric images and known spectroscopic redshifts from 0 < z < 4. We find that the performance is much improved when using images compared to photometry, with the BCNN achieving 0.098 rms error, a standard outlier rate of 3.9%, a 3σ outlier rate of 4.5%, and a bias of 0.0007. The performance drops significantly beyond z > 1.5 due to the relative lack of training data beyond those redshifts. This investigation demonstrates the power of using images directly and we advocate that future photo-z analysis of large-scale surveys include galaxy images.

A Real-World Study on Adverse Reactions of Belimumab Based on the FDA Adverse Event Reporting System Database.

This investigation leverages data derived from the United States Food and Drug Administration Adverse Event Reporting (FAERS) to real-world adverse reactions associated with Belimumab, with the intention of providing guidance for safe clinical pharmacotherapy. Data encompassing adverse drug event (ADE) reports relating to Belimumab from Q1 2011 to Q4 2023 within the FAERS were extracted and analyzed using methodologies such as the reporting odds ratio (ROR), proportional reporting ratio (PRR), Bayesian Confidence Propagation Neural Network (BCPNN), and Multi-item Gamma Poisson Shrinker (MGPS). The study identified a total of 19 825 ADE reports where Belimumab was the primary suspect medication, with the United States constituting the majority of reporting countries (16 312 cases, or 82.28%). Patients aged 18 to 64.9 years accounted for the largest demographic (36.29%), while the proportion of female patients (77.91%) significantly surpassed that of male patients (5.03%). The analysis uncovered 184 unique Preferred Terms (PTs) across 21 System Organ Classes (SOCs). Following selection through ROR, the SOC signal strength was prioritized as follows: Systemic disorders and administration site conditions, infections and infestations, a variety of musculoskeletal and connective tissue disorders, and conditions related to pregnancy, puerperium, and the perinatal period. The top five PTs for ADE reports not included in the product's labeling were hypersensitivity reactions, immunosuppression, non-vascular diseases, herpes virus infections, and Sjögren's syndrome. The top five PTs for ADE signal strength not included in the labeling were disseminated cutaneous herpes zoster, herpes zoster meningitis, onycholysis, cyclothymic disorder, and mixed connective tissue disease. Based on pharmacovigilance research utilizing the FAERS database, it is recommended that clinical monitoring of Bevacizumab should be intensified to support effective pharmaceutical care and ensure rational clinical medication use.

Reliable Prediction of Solar Photovoltaic Power and Module Efficiency using Bayesian Surrogate Assisted Explainable Data-Driven Model

This study proposes a Bayesian surrogate-driven explainable deep neural network model to predict and interpret the module efficiency and maximum output power of three commercially available photovoltaic modules: monocrystalline silicon, polycrystalline silicon, and amorphous silicon during the winter season. In addition, the influence of the photovoltaic material and ambient conditions on the predicted power and module efficiency is investigated. The model inputs include solar irradiance, module material (monocrystalline, polycrystalline, and amorphous silicon), season of the year (months), and time of day (morning to evening). Experiments were conducted on these three modules during the winter using data from Rawalpindi, Pakistan. These data were then fed into the deep neural network model to train and yield predictions. Several preprocessing techniques such as logarithmic transformation, square root, cube root, reciprocal, and exponential transformations are employed to improve the linearity of distributed data. For hyper-parameters optimization, Gaussian process, gradient boost regression trees, and random forest are used. The results show that the optimal deep neural network using random forest surrogate model along with square root and exponential transformation is able to predict the maximum power output and module efficiency of the considered photovoltaic modules for the entire investigated period with a correlation coefficient, R2 = 0.998. The least accurate model (R2=0.991) is a Gaussian process using simple data distribution. These results hold valuable insights for predicting and optimizing the performance of other solar photovoltaic modules in various climates and different seasons of the year.

Bayesian polynomial neural networks and polynomial neural ordinary differential equations.

Symbolic regression with polynomial neural networks and polynomial neural ordinary differential equations (ODEs) are two recent and powerful approaches for equation recovery of many science and engineering problems. However, these methods provide point estimates for the model parameters and are currently unable to accommodate noisy data. We address this challenge by developing and validating the following Bayesian inference methods: the Laplace approximation, Markov Chain Monte Carlo (MCMC) sampling methods, and variational inference. We have found the Laplace approximation to be the best method for this class of problems. Our work can be easily extended to the broader class of symbolic neural networks to which the polynomial neural network belongs.

Graph neural network-based multiscale thermal modeling for heterogeneous materials with complex structures

This study presents a novel GNN-based model for multiscale thermal characterization of heterogeneous materials with complex microstructures. It incorporates relationships between composition, structure, and thermal transport across multiple scales, leveraging the structure of GNNs. Results demonstrate enhanced performance of the GNN-based model compared with existing methods. Specifically, the GNN model improved thermal conductivity prediction with a mean absolute error of 0.18 W/mK (15 % improvement over the Bayesian neural network-based model) and provided a 30x speedup in computation time. The model successfully connected atomic-scale interactions to macroscale properties, achieving less than 5 % error in predicting effective thermal conductivity from one scale to another. Furthermore, it demonstrated the ability to capture nonlinear thermal transport phenomena with 92 % accuracy. The model also improved the prediction of the transient thermal responses by 20 %, accurately capturing the time-varying behaviour of materials. This work also proved the model’s ability to handle new material systems with transfer learning, achieving 88 % accuracy. Finally, Bayesian neural networks integrated into the model provided uncertainty quantification, with 95 % of experimental measurements falling within the predicted bounds. This work demonstrated the ability of the framework to handle complex microstructures, nonlinear phenomena, and transient responses while maintaining a computationally efficient model that can enable real-time applications in design and optimization needed in modern materials innovation. It represents a powerful tool for data-driven multiscale mechanics that could accelerate materials research and development for the aerospace, electronics cooling, and additive manufacturing industries.

Vehicle platoon safety considering the vehicle cut-in: A coupled reinforcement learning and model predictive control approach

A longitudinal platoon control method based on Twin Delayed Deep Deterministic Policy Gradient (TD3) and Model Predictive Control (MPC) is proposed to solve the problems of low following efficiency and system instability in longitudinal platoon control. Firstly, Dynamic Bayesian Network (DBN) and Long Short-Term Memory (LSTM) network are introduced to identify the driving behavior of bystanders and derive the MPC objective constraint function containing three indicators of following, comfort and fuel consumption according to the platoon dynamics equation. Secondly, the system prediction model and cost function are introduced into the action-critic network of TD3 to solve the problem of no model training in the traditional TD3 algorithm and to speed up the training speed and accuracy of the network. On this basis, a Bellman equation is proposed to calculate the time-domain difference error and the expected loss function to solve the TD3 network overestimation problem. Finally, the joint simulation platform is built to simulate the platoon driving conditions and compare with the DDPG optimization algorithm and the traditional MPC algorithm respectively. The results show that the improved TD3-MPC algorithm satisfies the constraints and the spacing error is controlled within 0.3 m, and the vehicle speed changes more smoothly in the scenario of speed fluctuation in the front vehicle, and the ride comfort of the platoon is improved, and it has better robustness in the scenario of vehicle cut-in. The experimental results show that when the vehicle speeds are 20, 40, and 60 km/h, compared to MPC, the average spacing errors are reduced by 27.56%, 25.64%, and 28.04%, respectively, and compared to DDPG-MPC, the average spacing errors are reduced by 6.59%, 8.77%, and 7.86%, respectively.

Data-Driven Bayesian Risk Assessment of Factors Influencing the Severity of Marine Accidents in Port Areas

The most prevalent types of ship accidents in port areas are allisions, collisions, and groundings. A comprehensive risk assessment is needed to prevent and mitigate these accidents and their consequences. This study evaluates the risk of such accidents in port areas by presenting a model that elucidates the relationship between risk-identifying factors (RIFs) and accident severity. In this context, the RIFs are determined by analyszing the reports of 528 accidents that occurred in port areas between 1995 and 2023. Subsequently, the model is created by analysing the data derived from these reports using the Tree Augmented Naive Bayes (TAN) algorithm, through an analysis of the data derived from these reports using the Tree Augmented Naive Bayes (TAN) algorithm, which is an approach of the data-driven Bayesian network method. The findings of the study indicate that accident type, wind, ship age, and vessel type are the most influential factors in predicting the severity of accidents in port areas. It is thought that the model will assist port authorities in identifying operational risks contributing to accidents and that contribute to accidents and in formulating preventive regulations.

Immune checkpoint inhibitor combinations for patients with advanced endometrial cancer: a network meta-analysis and cost-utility analysis

BackgroundImmune checkpoint inhibitor combinations show significant survival advantages compared with chemotherapy for patients with advanced endometrial cancer.ObjectiveTo compare the efficacy, safety, and cost-effectiveness of different immunotherapy combinations for clinician and...

8007 The Effect of Glucagon-Like Peptide-1 Receptor Agonists in Reducing Major Adverse Cardiovascular Events in Patients With Type 2 Diabetes Mellitus: A Bayesian Network Meta-Analysis

8007 The Effect of Glucagon-Like Peptide-1 Receptor Agonists in Reducing Major Adverse Cardiovascular Events in Patients With Type 2 Diabetes Mellitus: A Bayesian Network Meta-Analysis

Road Traffic Safety Status Analysis and Prediction Based on Dynamic Bayesian Network

Abstract Dynamic Bayesian networks can effectively capture dynamic changes and uncertainty relationships in data. Conventional prediction methods do not consider the temporal characteristics between traffic flow sequences, which affects prediction accuracy. This article proposes a method for analyzing and predicting road traffic safety status based on DBN. Firstly, data matching is performed according to the “case-control” sample structure of the matching formula to minimize the influence of other factors on the modeling of traffic safety status; Secondly, the random forest model is applied to analyze and extract the variable with the highest correlation coefficient as the input variable for the traffic safety status prediction model; Then, a DBN prediction model is established using matched accident traffic flow and non-accident traffic flow sample data; Finally, by analyzing the effectiveness evaluation indicators of the model, multiple prediction results showed that the overall prediction accuracy of the DBN method was over 80%.

Classification-based consensus model considering quantum interference in linguistic distribution environment

Multi-attribute group classification (MAGC) is widely used in dealing with a variety of practical problems. However, cognitive biases, resulting from factors such as irrational behaviors, epistemic uncertainty, and interference phenomena, may critically affect the consensus reaching process (CRP) of MAGC. To eliminate such influences, this study proposes an ordinal classification consensus model in a linguistic environment based on quantum probability theory (QPT), with the aim of classifying the alternatives into several predefined ordinal classifications accepted by a majority of experts. First, the linguistic distribution assessments (LDAs), which can depict the quantitative distribution and qualitative vagueness, are used to express the uncertain preference of experts. Second, the LDAs best-worst method (BWM) and LDAs maximizing deviation model (MDM) are developed to identify the subjective and objective weight of attributes, and then these two types of attributes weight are integrated by quantum-like Bayesian networks (QLBN). Third, a maximum consensus model (MCM), maximizing the consensus level of ordinal classification results of experts, is provided to determine the parameters of QLBN. When the consensus level is lower than the predetermined threshold, a minimum adjustment distance model (MADM) is designed to guide the preference-modification of experts to derive the consensus classification of alternatives. Finally, a risk assessment case study is conducted to verify the feasibility and effectiveness of the proposed method. Sensitivity analysis and comparison analysis are performed to verify the feasibility and effectiveness of the proposed method

An aircraft structural risk assessment method considering fatigue crack propagation based on fatigue damage diagnosis and prognosis

An aircraft structural risk assessment method based on fatigue damage diagnosis and prognosis has been developed, considering fatigue crack propagation. The process is divided into three stages: initial crack diagnosis, crack diagnosis, and prediction, utilizing Monte Carlo simulation. Using 2024 aluminum alloy specimens with central holes, the study indicates that in the initial crack diagnosis stage, an inspection standard with a Single Flight Probability of Failure (SFPOF) less than 10-7 and a threshold method enhances structural fatigue crack diagnosis. In the crack diagnosis and prediction stages, iterative updates using Gaussian Process Regression (GPR) within a Dynamic Bayesian Network (DBN) improve crack propagation prediction and risk assessment accuracy. The diagnostic interval significantly impacts SFPOF, with an optimized interval balancing accuracy and computation time. Simplified and precise K value calculation methods enhance efficiency and accuracy. The method reduces costs and improves risk assessment accuracy, providing new insights for SPHM-based aircraft structural risk assessment.

A Bayesian Network Inference Approach for Dynamic Risk Assessment Using Multisource-Based Information Fusion in an Interval Type-2 Fuzzy Set Environment

A Bayesian Network Inference Approach for Dynamic Risk Assessment Using Multisource-Based Information Fusion in an Interval Type-2 Fuzzy Set Environment

Deep Belief Network with Fuzzy Parameters and Its Membership Function Sensitivity Analysis

Over the last few years, deep belief networks (DBNs) have been extensively utilized for efficient and reliable performance in several complex systems. One critical factor contributing to the enhanced learning of the DBN layers is the handling of network parameters, such as weights and biases. The efficient training of these parameters significantly influences the overall enhanced performance of the DBN. However, the initialization of these parameters is often random, and the data samples are normally corrupted by unwanted noise. This causes the uncertainty to arise among weights and biases of the DBNs, which ultimately hinders the performance of the network. To address this challenge, we propose a novel DBN model with weights and biases represented using fuzzy sets. The approach systematically handles inherent uncertainties in parameters resulting in a more robust and reliable training process. We show the working of the proposed algorithm considering four widely used benchmark datasets such as: MNSIT, n-MNIST (MNIST with additive white Gaussian noise (AWGN) and MNIST with motion blur) and CIFAR-10. The experimental results show superiority of the proposed approach as compared to classical DBN in terms of robustness and enhanced performance. Moreover, it has the capability to produce equivalent results with a smaller number of nodes in the hidden layer; thus, reducing the computational complexity of the network architecture. Additionally, we also study the sensitivity analysis for stability and consistency by considering different membership functions to model the uncertain weights and biases. Further, we establish the statistical significance of the obtained results by conducting both one-way and Kruskal-Wallis analyses of variance tests.

Modeling CO2 loading capacity of triethanolamine (TEA) aqueous solutions via a deep learning approach

Capturing carbon dioxide (CO2) from natural gas is essential for reducing CO2 emissions as a greenhouse gas as well as increasing the gas heating value. Absorption of CO2 by amine solutions is a method that has seen a widespread use as a CO2 collection technique. In this research, advanced ML approaches are utilized to simulate the CO2 loading capacity of triethanolamine (TEA) aqueous solutions as a function of system temperature, CO2 partial pressure, and amine concentration in the aqueous phase. Deep neural network (DNN), Gaussian process regressor (GPR), deep belief network (DBN) and bagging regressor (BR) are the models that have been developed. The DNN model with the coefficient of determination (R2) of 0.9990 as well as the root mean square error (RMSE) of 0.0073 outperformed other models in terms of accuracy and validity. The R2 values of 0.9911, 0.9861, and 0.9745 for DBN, BR, and GPR, accordingly, showed that other models could likewise perform with high accuracy. Moreover, trend analysis of the results confirmed that the DNN method can correctly estimate the behavior of CO2 loading with variations in the input parameters. Furthermore, a sensitivity analysis showed that temperature has a decreasing effect on the CO2 loading, while amine concentration and CO2 partial pressure exhibited to have an incremental effect. Herein, the temperature had the greatest influence on the amount of CO2 loading. The final stage of the research was the leverage approach, which stated that about 99% of the data are statistically valid and the DNN model is reliable to be replaced by experimental approaches in the prediction of CO2 absorption into certain type of solutions.

Detection of foreign materials on Semen Ziziphi Spinosae using hyperspectral imaging technology coupled with convolutional neural networks

The efficacy of Semen Ziziphi Spinosae, a valued medicinal material, can be compromised by the presence of foreign materials, which may also worsen patient conditions and lead to complications. Traditional machine vision techniques often struggle to identify foreign entities that closely resemble Semen Ziziphi Spinosae in color and shape. This study investigates the application of hyperspectral imaging technology for the detection and differentiation of four types of foreign materials (moldy Semen Ziziphi Spinosae shell, wood, and stone) in processing industries. High-resolution hyperspectral imagery was subjected to image segmentation and background noise reduction to separate the contours of the samples within the images. The average spectral signatures of the regions of interest were extracted to construct models such as Convolutional Neural Networks (CNN), Support Vector Machines (SVMs), Mahalanobis distance, Deep Belief Networks, and Generative Adversarial Networks. These models were employed to extract spectral information indicative of foreign matter within the hyperspectral data to identify foreign materialss in Semen Ziziphi Spinosae. Subsequently, the classification outcomes were mapped onto the original hyperspectral images, enabling a visual representation of foreign matter within the seeds. The classification performance of alternative models was compared with that of the CNN model. The results demonstrate that a CNN trained with spectral information from hyperspectral data can accurately identify foreign materialss in Semen Ziziphi Spinosae. The model underwent 30 iterations, achieving an accuracy rate of 99% with a root mean square error (RMSE) of 0.62 for the validation detection, and an accuracy rate of 98% with an RMSE of 0.62 for the validation set.

Threshold-optimized and features-fused semi-supervised domain adaptation method for rotating machinery fault diagnosis

In the field of intelligent fault diagnosis, domain adaptation (DA) technology achieves significant breakthroughs, particularly in reducing reliance on large volumes of labeled samples. Despite these advancements, challenges persist when unlabeled data do not accurately represent actual application scenarios. Additionally, the impact of pseudo-labels on conditional domain adaptation raises concerns. To overcome the above challenges, a novel DA approach based on chaos sparrow search algorithm (CSSOA) optimized threshold parameters and feature fusion deep belief network (DBN) is proposed, named CSS-DADBN. Firstly, this method, by integrating pseudo-label updating with semi-supervised domain adaptation (SSDA) and employing confidence and entropy threshold parameters as corrective rules for pseudo-label filtering, along with the introduction of iterative conditions as an additional selection criterion, effectively alleviates the aforementioned issues. Furthermore, combining the feature extraction capabilities of DBN with a domain feature fusion strategy significantly enhances cross-domain feature learning, thereby substantially improving diagnostic accuracy. Ultimately, to validate the effectiveness and practicality of the CSS-DADBN method, a series of experiments conducted on the PT700 and Case Western Reserve University (CWRU) rolling bearing test platform clearly demonstrate its utility and efficiency in intelligent fault diagnosis.

ScCrab: A Reference-Guided Cancer Cell Identification Method based on Bayesian Neural Networks.

Cancer is a significant global public health concern, where early detection can greatly enhance curative outcomes. Therefore, the identification of cancer cells holds significant importance as the primary method for cancer diagnosis. The advancement of single-cell RNA sequencing (scRNA-seq) technology has made it possible to address the problem of cancer cell identification at the single-cell level more efficiently with computational methods, as opposed to the time-consuming and less reproducible manual identification methods. However, existing computational methods have shown suboptimal identification performance and a lack of capability to incorporate external reference data as prior information. Here, we propose scCrab, a reference-guided automatic cancer cell identification method, which performs ensemble learning based on a Bayesian neural network (BNN) with multi-head self-attention mechanisms and a linear regression model. Through a series of experiments on various datasets, we systematically validated the superior performance of scCrab in both intra- and inter-dataset predictions. Besides, we demonstrated the robustness of scCrab to dropout rate and sample size, and conducted ablation experiments to investigate the contributions of each component in scCrab. Furthermore, as a dedicated model for cancer cell identification, scCrab effectively captures cancer-related biological significance during the identification process.