Article
Chemistry, Physical
Yunli Wang, Sijia Wang, Cyrille Deces-Petit
Summary: The ability to evaluate measurement error at hydrogen refueling stations is crucial for the sustainability of the hydrogen vehicle industry. This study focuses on estimating measurement accuracy using data collected from real operation of hydrogen refueling stations. The proposed method uses a Bayesian non-parametric approach called Dirichlet process mixture models, which effectively captures the complex structure of the data and estimates the probability distribution of measurement uncertainty.
INTERNATIONAL JOURNAL OF HYDROGEN ENERGY
(2022)
Article
Engineering, Civil
Weixing Liu, Zunqian Zhang, Yunjie Bai, Yikai Liu, Aimin Yang, Jie Li
Summary: This study proposes a sustainable prediction model for passenger car sales based on DeepFM. By using feature engineering to expand explicit features, a multilayer neural network to extract implicit features, and a Bayesian neural network to infer predictive values, combined with a factorization machine to consider cross information, the model’s expressive and predictive power for unknown data is improved. Experimental results verify the superiority of the proposed method.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Engineering, Mechanical
Xu-Yang Cao, De-Cheng Feng, Michael Beer
Summary: With the development of performance-based earthquake engineering, the risk-informed assessment framework has gained recognition worldwide, particularly the probability seismic fragility analysis. Researchers are exploring non-parametric approaches to express intrinsic fragility without distribution assumptions, while also considering calculation efficiency and non-stationary stochastic responses. This paper proposes a kernel density estimation-based non-parametric cloud approach for efficient seismic fragility estimation and demonstrates its effectiveness through an application example. The findings provide insights for the development of non-parametric seismic fragility approaches.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Statistics & Probability
Ran Bi, Peng Liu
Summary: The molecular mechanism of heterosis, the superior performance of hybrid offspring over their inbred parents, is not well studied. In this study, the authors propose a statistical model and a powerful test to detect genes exhibiting heterosis at the gene expression level. The method, based on a Bayesian framework and Markov Chain Monte Carlo sampling, outperforms other methods in detecting gene expression heterosis according to simulation results.
JOURNAL OF APPLIED STATISTICS
(2023)
Article
Thermodynamics
Freddy Houndekindo, Taha B. M. J. Ouarda
Summary: Statistical methods are useful in estimating wind resources and identifying locations for further investigation. This study proposes a non-parametric approach to map wind speed distribution, which is more flexible than existing methods. The results show that this approach performs better in regions with high wind regime variability.
ENERGY CONVERSION AND MANAGEMENT
(2023)
Article
Computer Science, Artificial Intelligence
Hajer Salem, Moamar Sayed-Mouchaweh, Moncef Tagina
Summary: Infinite Factorial Hidden Markov Model (iFHMM) is an extension of the Factorial Hidden Markov Model for Non-Intrusive Load Monitoring (NILM), inferring the number of appliances in households and adjusting its model complexity accordingly. To overcome the challenges faced by the original model, a new version, iFHMMCC, is introduced with contextual features constraints to enhance accuracy and computational efficiency.
Article
Computer Science, Artificial Intelligence
Xiang Yang, Peter Meer, Jonathan Meer
Summary: Most robust estimators require tuning parameters for practical applications, but MISRE tackles this issue by processing different structures independently. With inlier structures listed first during data processing and ordering, the inlier/outlier classification becomes straightforward. MISRE performs similarly to RANSAC-type algorithms when inlier noises are similar.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Engineering, Multidisciplinary
Mainak Bhattacharyya, Pierre Feissel
Summary: The research aims to determine material parameters through full field measurements of kinematic data from digital image correlation, using both deterministic and probabilistic approaches. The deterministic inverse problem involves an optimal control method that does not require complete knowledge of boundary conditions and measurement data. The probabilistic framework incorporates this method into a Bayesian inference framework, using a Markov chain Monte Carlo sampling method to obtain the posterior probability density function, supported by a radial basis function network for efficient sampling.
INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING
(2023)
Article
Computer Science, Software Engineering
Miguel Angel Mendza-Lugo, Oswaldo Morales-Napoles
Summary: The new update (version 1.3) of the BANSHEE-A MATLAB toolbox for Non-Parametric Bayesian Networks introduces functional enhancements. The additions include quantifying fully parametric Bayesian Networks, generating a user-defined sample size of (un)conditional samples, conditioning sampling based on samples, and selecting different sample sizes for model validation tests using the Hellinger distance. The update also incorporates the use of t copula in the Cramer-Von Mises test and provides detailed examples demonstrating the new features.
Article
Business
Sachin S. Kamble, Angappa Gunasekaran, Vikas Kumar, Amine Belhadi, Cyril Foropon
Summary: This paper aims to provide a decision support system for managers to predict an organization's probability of successful blockchain adoption. Factors such as competitor pressure, partner readiness, perceived usefulness, and perceived ease of use were found to have the most influence on blockchain adoption behavior. The use of a Bayesian network analysis in developing a predictive decision support system can help decision-makers assess their adoption probability and develop future adoption strategies.
TECHNOLOGICAL FORECASTING AND SOCIAL CHANGE
(2021)
Article
Engineering, Biomedical
Samah Khawaled, Moti Freiman
Summary: This study introduces the NPBDREG method for DNN-based image registration, which accurately estimates the uncertainty of registration and improves the registration accuracy and smoothness. It also demonstrates superior generalization capability when handling corrupted data.
COMPUTERIZED MEDICAL IMAGING AND GRAPHICS
(2022)
Article
Computer Science, Theory & Methods
Michael Minyi Zhang, Sinead A. Williamson, Fernando Perez-Cruz
Summary: This study presents a method to accelerate latent variable model inference by proposing feature locations based on the data, allowing for quick convergence and parallelization.
STATISTICS AND COMPUTING
(2022)
Article
Hospitality, Leisure, Sport & Tourism
A. George Assaf, Mike Tsionas, Florian Kock, Alexander Josiassen
Summary: This paper introduces a new Bayesian non-parametric stochastic frontier model that addresses the endogeneity problem and relaxes assumptions concerning functional form and distributional properties. The model is shown to outperform its parametric counterpart in critical diagnostic tests. Efficiency results obtained from a unique sample of US hotels operating within competitive clusters are used to analyze performance spillover effects and their implications for future co-location strategies.
ANNALS OF TOURISM RESEARCH
(2021)
Article
Environmental Sciences
Chingka Kalai, Arpita Mondal
Summary: Regional frequency analysis (RFA) is a useful tool for flood estimation and prediction in ungauged basins. In areas with strong climate seasonality, estimation of annual peakflow seasonality is important for effective water infrastructure management and timely flood prevention. This study proposes a novel RFA framework based on circular statistics for estimating annual peakflow seasonality. The framework includes attribute selection, formation of homogeneous regions, homogeneity tests for directional data, and probabilistic prediction using non-parametric regional circular density. The proposed approach is demonstrated on synthetic and real-world cases, showing good performance.
WATER RESOURCES RESEARCH
(2022)
Article
Computer Science, Artificial Intelligence
Ziyue Wu, Xi Chen, Zhaoxing Gao
Summary: This article proposes a novel Bayesian model called PoissonGP for forecasting online product sales. The model is able to capture complex patterns in sales data and manage distribution shifts caused by changes in long-run sales. Experimental results demonstrate that PoissonGP outperforms existing approaches, making it a promising tool for ecommerce platforms.
DECISION SUPPORT SYSTEMS
(2023)