Fast nuclide identification method based on hybrid dynamic Bayesian network.

A hybrid Dynamic Bayesian network method for failure prediction of a lock mechanism

Bayesian networks for the static as well as for the dynamic cases have been the subject of a great deal of theoretical analysis and practical inference approximations in the research community of artificial intelligence, machine learning and pattern recognition. After exploring the quite well known theory of discrete and continuous Bayesian networks, we introduce an almost instant reasoning scheme to the hybrid Bayesian networks. In addition to illustrate the similarities of the dynamic Bayesian networks (DBN) and the Kalman filter, we present a computationally efficient approach for the inference problem of hybrid dynamic Bayesian networks (HDBN). The proposed method is based on the separations of the dynamic and static nodes, and following hypercubic partitions via the Decision tree algorithm (DT). Experiments show that with high statistical confidence the novel algorithm used in the HDBN performs favorably in the tradeoffs of computational complexities and accuracy performance when compared to Junction tree and Gaussian mixture models on the task of classifications.

In this paper we describe the use of hybrid dynamic Bayesian networks (HDBNs) to model the operational risk faced by financial institutions in terms of economic capital. We describe a methodology for modeling financial losses resulting from intentional or accidental events and characterize these by their ability to evade controls and which ultimately lead to increasingly severe financial consequences. The approach presented here focuses on modeling the causes and effects of loss events using a dynamic Bayesian network model based on interactions between failure modes and controls. To calculate the value-at-risk for total losses we apply a new state-of-the-art hybrid Bayesian network algorithm, called dynamic discretization. The algorithm approximates the continuous loss distribution functions required for each loss event at each point in time and is used to aggregate across loss types. In order to illustrate the natural match between the model and the underlying process, including the causal complexity underlying known and possible severe operational risk losses, we apply the generalized model to a financial trading example: rogue trading. We conclude that the statistical properties of the model have the potential to explain recent large-scale loss events and offer improved means of loss prediction.

BackgroundMalaria is a major cause of death in children under five years old in low- and middle-income countries such as Malawi. Accurate diagnosis and management of malaria can help reduce the global burden of childhood morbidity and mortality. Trained healthcare workers in rural health centers manage malaria with limited supplies of malarial diagnostic tests and drugs for treatment. A clinical decision support system that integrates predictive models to provide an accurate prediction of malaria based on clinical features could aid healthcare workers in the judicious use of testing and treatment. We developed Bayesian network (BN) models to predict the probability of malaria from clinical features and an illustrative decision tree to model the decision to use or not use a malaria rapid diagnostic test (mRDT).MethodsWe developed two BN models to predict malaria from a dataset of outpatient encounters of children in Malawi. The first BN model was created manually with expert knowledge, and the second model was derived using an automated method. The performance of the BN models was compared to other statistical models on a range of performance metrics at multiple thresholds. We developed a decision tree that integrates predictions with the costs of mRDT and a course of recommended treatment.ResultsThe manually created BN model achieved an area under the ROC curve (AUC) equal to 0.60 which was statistically significantly higher than the other models. At the optimal threshold for classification, the manual BN model had sensitivity and specificity of 0.74 and 0.42 respectively, and the automated BN model had sensitivity and specificity of 0.45 and 0.68 respectively. The balanced accuracy values were similar across all the models. Sensitivity analysis of the decision tree showed that for values of probability of malaria below 0.04 and above 0.40, the preferred decision that minimizes expected costs is not to perform mRDT.ConclusionIn resource-constrained settings, judicious use of mRDT is important. Predictive models in combination with decision analysis can provide personalized guidance on when to use mRDT in the management of childhood malaria. BN models can be efficiently derived from data to support clinical decision making.

Early warning method for overseas natural gas pipeline accidents based on FDOOBN under severe environmental conditions

A dynamic Bayesian network model for real-time crash prediction using traffic speed conditions data

The Usefulness of Bayesian Network in Assessing the Risk of Triple-Negative Breast Cancer

A Bayesian network is a powerful tool for modeling relations among a large number of random variables. Therefore the Bayesian network has received considerable attention from the studies of gene network estimation using microarray gene expression data. Imoto et al. [1, 2] proposed a Bayesian network and nonparametric regression model for capturing nonlinear relations between genes from the continuous gene expression data. However, a Bayesian network still has a problem that it cannot construct cyclic regulations, while real gene networks have cyclic regulations. For a solution of this problem, in this paper, we propose a dynamic Bayesian network and nonparametric regression model for estimating a gene network with cyclic regulations from time series microarray data. We also derive a criterion for selecting a network from Bayes approach. The effectiveness of our method is displayed though the analysis of the Saccharomyces cerevisiae gene expression data.

A hybrid intelligent model for reservoir production and associated dynamic risks

In industrial processes, complex faults or multi-faults can cause more comprehensive or wide-spread phenomenon than single-faults, leading to severe consequences. Thus complete fault diagnosis techniques should be able to handle multi-fault as well as single-fault cases although multi-fault cases have relatively low probability and high complexity. In order to describe the conditional relationship between process variables and known faults, a Bayesian network can be employed, where process variables are continuous while faults are discrete; this case is called hybrid. In addition, time factor or dynamics should also be included, leading to the hybrid dynamic Bayesian network (HDBN) framework, which has been used to describe and monitor dynamic systems. Under this framework, we describe fault diagnosis problems as a HDBN inference problem and propose an algorithm based on time iteration. A simulated 5-sink system and the Tennessee Eastman Process (TEP) are given to illustrate and validate the proposed methodology and some practical issues are discussed. The performance of the algorithm when treating the TEP fault diagnosis indicates a balance between isolation accuracy and computational complexity.

In this article, we develop a dynamic Bayesian network (DBN) model to measure 3D visual fatigue. As far as our information goes, this is the first adaptation of a DBN structure-based probabilistic framework for inferring the 3D viewer's state of visual fatigue. Our measurement focuses on the interdependencies between each factor and the phenomena of visual fatigue in stereoscopy. Specifically, the implementation of DBN with using multiple features (e.g. contextual, contactless and contact physiological features) and dynamic factor provides a systematic scheme to evaluate 3D visual fatigue. In contrast to measurement results between the mean opinion score (MOS) and Bayesian network model (with static Bayesian network and DBN), the visual fatigue in stereoscopy at time slice t is influenced by a dynamic factor (time slice t-1). In the presence of dynamic factors (time slice t-1), our proposed measuring scheme based on DBN is more comprehensive. (i) We cover more features for inferring the visual fatigue, more reliably and accurately; (ii) at different time slices, the dynamic factor features are significant for inferring the visual fatigue state of stereoscopy.

Classic statistical and machine learning models such as support vector machines (SVMs) can be used to predict cancer outcome, but often only perform well if all the input variables are known, which is unlikely in the medical domain. Bayesian network (BN) models have a natural ability to reason under uncertainty and might handle missing data better. In this study, the authors hypothesize that a BN model can predict two-year survival in non-small cell lung cancer (NSCLC) patients as accurately as SVM, but will predict survival more accurately when data are missing. A BN and SVM model were trained on 322 inoperable NSCLC patients treated with radiotherapy from Maastricht and validated in three independent data sets of 35, 47, and 33 patients from Ghent, Leuven, and Toronto. Missing variables occurred in the data set with only 37, 28, and 24 patients having a complete data set. The BN model structure and parameter learning identified gross tumor volume size, performance status, and number of positive lymph nodes on a PET as prognostic factors for two-year survival. When validated in the full validation set of Ghent, Leuven, and Toronto, the BN model had an AUC of 0.77, 0.72, and 0.70, respectively. A SVM model based on the same variables had an overall worse performance (AUC 0.71, 0.68, and 0.69) especially in the Ghent set, which had the highest percentage of missing the important GTV size data. When only patients with complete data sets were considered, the BN and SVM model performed more alike. Within the limitations of this study, the hypothesis is supported that BN models are better at handling missing data than SVM models and are therefore more suitable for the medical domain. Future works have to focus on improving the BN performance by including more patients, more variables, and more diversity.

This paper is a revision of a paper presented at the SPIE conference on Signal Processing, Senior Fusion, and Target Recognition XII, Aug. 2004, Orlando, Florida. The paper presented there appears (unrefereed) in SPIE Proceedings Vol. 5429. Bayesian networks for static as well as for dynamic cases have been the subject of a great deal of theoretical analysis and practical inference-algorithm development in the research community of artificial intelligence, machine learning, and pattern recognition. After summarizing the well-known theory of discrete and continuous Bayesian networks, we introduce an efficient reasoning scheme into hybrid Bayesian networks. In addition to illustrating the similarities between the dynamic Bayesian networks and the Kalman filter, we present a computationally efficient approach for the inference problem of hybrid dynamic Bayesian networks (HDBNs). The proposed method is based on the separation of the dynamic and static nodes, and subsequent hypercubic partitions via the decision tree algorithm. Experiments show that with high statistical confidence the novel algorithm used in the HDBN performs favorably in the trade-offs of computational complexity and accuracy performance, compared to other exact and approximate methods for applications with uncertainty in a dynamic system.

The management of environmental problems is multifaceted, requiring varied and sometimes conflicting objectives and perspectives to be considered. Bayesian network (BN) modeling facilitates the integration of information from diverse sources and is well suited to tackling the management challenges of complex environmental problems. However, combining several perspectives in one model can lead to large, unwieldy BNs that are difficult to maintain and understand. Conversely, an oversimplified model may lead to an unrealistic representation of the environmental problem. Environmental managers require the current research and available knowledge about an environmental problem of interest to be consolidated in a meaningful way, thereby enabling the assessment of potential impacts and different courses of action. Previous investigations of the environmental problem of interest may have already resulted in the construction of several disparate ecological models. On the other hand, the opportunity may exist to initiate this modeling. In the first instance, the challenge is to integrate existing models and to merge the information and perspectives from these models. In the second instance, the challenge is to include different aspects of the environmental problem incorporating both the scientific and management requirements. Although the paths leading to the combined model may differ for these 2 situations, the common objective is to design an integrated model that captures the available information and research, yet is simple to maintain, expand, and refine. BN modeling is typically an iterative process, and we describe a heuristic method, the iterative Bayesian network development cycle (IBNDC), for the development of integrated BN models that are suitable for both situations outlined above. The IBNDC approach facilitates object-oriented BN (OOBN) modeling, arguably viewed as the next logical step in adaptive management modeling, and that embraces iterative development. The benefits of OOBN modeling in the environmental community have not yet been fully realized in environmental management research. The IBNDC approach to BN modeling is described in the context of 2 case studies. The first is the initiation of blooms of Lyngbya majuscula, a blue-green algae, in Deception Bay, Australia where 3 existing models are being integrated, and the second case study is the viability of the free-ranging cheetah (Acinonyx jubatus) population in Namibia where an integrated OOBN model is created consisting of 3 independent subnetworks, each describing a particular aspect of free-ranging cheetah population conservation.

A comparison between discrete and continuous time Bayesian networks in learning from clinical time series data with irregularity.