Overview of Bayesian Belief Network

PurposeExpert systems have come to the forefront in the modeling of problems. One of the major problems facing the expert system designers is to develop an accurate knowledge base and a meaningful model of uncertainty associated with complex models. Decision-making is based on knowledge, and decision system support needs a knowledge base as well. An adequate knowledge acquisition (KA) process leads to accurate knowledge and improves the decision-making process. To manage the risk of a medical service (twin pregnancy in this case) a knowledge management system was created. The captured knowledge may be associated with an uncertainty. This study aims to introduce a method for evaluating the reliability of a tacit KA model. It assisted engineering managers in assessing and prioritizing risks. The study tried to use this method in risk management and new case in the health domain.Design/methodology/approachIn this study, relevant variables were identified in the knowledge management literature reviews and the domain of expertise management. They are validated by a group of domain experts. Kendall’s W indicator was used to assess the degree of consensus. On the basis of combined cognitive maps, a cognitive network was constructed. Using Bayesian belief networks and fuzzy cognitive maps, an uncertainty assessment method of tacit KA was introduced. To help managers focus on major variables, a sensitivity analysis was conducted. Reliability of model was calculated for optimistic and pessimistic values. The applicability and efficacy of the proposed method were verified and validated with data from a medical university.FindingsResults show that tacit KA uncertainty can be defined by independent variables, including environmental factors, personality and acquisition process factors. The reliability value shows the accuracy of the captured knowledge and the effectiveness of the acquisition process. The proposed uncertainty assessment method provides the reliability value of the acquisition model for knowledge engineers, so it can be used to implement the project and prevent failures in vital factors through necessary actions. If there is not a satisficed level of reliability, the KA project reliability can be improved by risk factors. The sensitivity analysis can help to select proper factors based on the resources. This approach mitigated some of the disadvantages of other risk evaluation methods.Originality/valueThe contribution of this study is to combine the uncertainty assessment with tacit KA based on fuzzy cognitive maps and the Bayesian belief networks approach. This approach used the capabilities of both narrative and computational approaches.

Bayesian belief networks as a meta-modelling tool in integrated river basin management — Pros and cons in evaluating nutrient abatement decisions under uncertainty in a Norwegian river basin

The Kilombero Valley floodplain in Tanzania is a major agricultural area. Government initiatives and projects supported by international funding have long sought to boost productivity. Due to increasing population pressure, smallholder farmers are forced to increase their output. Nevertheless, the level of intensification is still lower than what is considered necessary to increase production and support smallholder livelihoods significantly. This article aims to better understand farmers’ intensification choices and their interdependent determinants. We propose a novel modeling approach for identifying determinants of intensification and their interrelationships by combining a Bayesian belief network (BBN), experimental design, and multivariate regression trees. Our approach complements existing lower‐dimensional statistical models by considering uncertainty and providing an easily updatable model structure. The BBN is constructed and calibrated using data from a survey of 304 farm households. Our findings show how the data‐driven BBN approach can be used to identify variables that influence farmers’ decision to choose one technique over another. Furthermore, the most important drivers vary widely, depending on the intensification options being considered.

The ability to predict spatial variation in biodiversity is a long-standing but elusive objective of landscape ecology. It depends on a detailed understanding of relationships between landscape and patch structure and taxonomic richness, and accurate spatial modelling. Complex heterogeneous environments such as cities pose particular challenges, as well as heightened relevance, given the increasing rate of urbanisation globally. Here we use a GIS-linked Bayesian Belief Network approach to test whether landscape and patch structural characteristics (including vegetation height, green-space patch size and their connectivity) drive measured taxonomic richness of numerous invertebrate, plant, and avian groups. We find that modelled richness is typically higher in larger and better-connected green-spaces with taller vegetation, indicative of more complex vegetation structure and consistent with the principle of ‘bigger, better, and more joined up’. Assessing the relative importance of these variables indicates that vegetation height is the most influential in determining richness for a majority of taxa. There is variation, however, between taxonomic groups in the relationships between richness and landscape structural characteristics, and the sensitivity of these relationships to particular predictors. Consequently, despite some broad commonalities, there will be trade-offs between different taxonomic groups when designing urban landscapes to maximise biodiversity. This research demonstrates the feasibility of using a GIS-coupled Bayesian Belief Network approach to model biodiversity at fine spatial scales in complex landscapes where current data and appropriate modelling approaches are lacking, and our findings have important implications for ecologists, conservationists and planners.

Amodel based on Bayesian belief network is put forward to solve the innovation performance problems of the urban subway construction safety in China. After a systematic analyzing of the influence factors on the technological innovation performance, 13 variables were chosen in the light of implementation factors, interference factors, mediation factors and control factors, by taking the application of building information model as a case study. The causal relationship between these variables was constructed by the cause and effect diagram of experts' prior knowledge, combined with the correlation analysis method to determine the Bayesian belief network system structure. Sample data was got through a structural questionnaire survey. It was fit by NETICA software to get the probability density of each node. The application results show that this method can achieve a certain quantitative analysis of the innovation performance of the subway construction safety technology, and has better compliance compared with actual engineering situation.

Agricultural intensification often has complex effects on a wide range of environmental and economic values, presenting planners with challenging decisions for optimising sustainable benefits. Bayesian Belief Networks (BBNs) can be used as a decision-support tool for evaluating the influence of development scenarios across a range of values. A BBN was developed to guide decisions on water abstraction and irrigation-driven land use intensification in the Hurunui River catchment, New Zealand. The BBN examines the combined effects of different irrigation water sources and four land development scenarios, with and without a suite of on-farm mitigations, on ground and surface water quality, key socioeconomic values (i.e. farm earnings and jobs, and contribution to regional gross domestic production (GDP)) and aquatic values (i.e. salmon, birds, waterscape, contact recreation, periphyton and invertebrates). It predicts high farm earnings, jobs and regional GDP with 150% increase in irrigated area, but a range of positive and negative aquatic environmental outcomes, depending on the location of water storage dams and the application of a suite of on-farm mitigations. This BBN synthesis of a complex system enhanced the ability to include aquatic values alongside economic and social values in land-use and water resource planning and decision making, and has influenced objective setting in Hurunui planning processes.

Estimating the impact of health systems factors on antimicrobial resistance in priority pathogens

The Kunming-Montreal Global Biodiversity Framework set an ambitious target of safeguarding 30% of the Earth’s land and sea through well-managed Protected Areas (PAs) by 2030. Despite 196 countries committing to expand PA coverage, a lack of evidence on existing PA effectiveness and limited data on human-ecosystem interactions pose formidable challenges to implementation and prioritising conservation funding. Within this context, this paper aims to test a simple and user-friendly model, which is based on Bayesian Belief Networks, in evaluating the management effectiveness of Cu Lao Cham Marine Protected Area (CLC-MPA) in Vietnam. It focuses on assessing whether the MPA is effective in achieving its conservation objectives and if the target resources can endure the current pace of economic development. The study underscores the importance of prioritising eco-tourism and sustainable local economic models amid the intense forces of mass tourism development. It argues that tourism carrying capacity for CLC should be estimated at the outset of sustainable tourism development plan. Besides, there is a need for an integrated approach to MPA management to include legal provisions to address the impacts of tourism and pollution. Also, given the increasing pressure from incoming fishers, it is essential for increasing trans-provincial coordination for better enforcement of MPA regulations.

This paper addresses the problem of online tracking and classification of multiple objects in an image sequence. Our proposed solution is to first track all objects in the scene without relying on object-specific prior knowledge, which in other systems can take the form of hand-crafted features or user-based track initialization. We then classify the tracked objects with a fast-learning image classifier, that is based on a shallow convolutional neural network architecture and demonstrate that object recognition improves when this is combined with object state information from the tracking algorithm. We argue that by transferring the use of prior knowledge from the detection and tracking stages to the classification stage, we can design a robust, general purpose object recognition system with the ability to detect and track a variety of object types. We describe our biologically inspired implementation, which adaptively learns the shape and motion of tracked objects, and apply it to the Neovision2 Tower benchmark data set, which contains multiple object types. An experimental evaluation demonstrates that our approach is competitive with the state-of-the-art video object recognition systems that do make use of object-specific prior knowledge in detection and tracking, while providing additional practical advantages by virtue of its generality.

In mountain forests of Central Europe, storm and snow breakage as well as bark beetles are the prevailing major disturbances. The complex interrelatedness between climate, disturbance agents, and forest management increases the need for an integrative approach explicitly addressing the multiple interactions between environmental changes, forest management, and disturbance agents to support forest resource managers in adaptive management. Empirical data with a comprehensive coverage for modelling the susceptibility of forests and the impact of disturbance agents are rare, thus making probabilistic models, based on expert knowledge, one of the few modelling approaches that are able to handle uncertainties due to the available information. Bayesian belief networks (BBNs) are a kind of probabilistic graphical model that has become very popular to practitioners and scientists mainly due to considerations of risk and uncertainties. In this contribution, we present a development methodology to define and parameterize BBNs based on expert elicitation and approximation. We modelled storm and bark beetle disturbances agents, analyzed effects of the development methodology on model structure, and evaluated behavior with stand data from Norway spruce (Picea abies (L.) Karst.) forests in southern Austria. The high vulnerability of the case study area according to different disturbance agents makes it particularly suitable for testing the BBN model.

Sepsis is a serious public health problem. sTREM-1 is a marker of inflammatory and infectious processes that has the potential to become a useful tool for predicting the evolution of sepsis. A prediction model for sepsis was constructed by combining sTREM-1, CRP, and a leukogram via a Bayesian network. A translational study carried out with 32 children with congenital heart disease who had undergone surgical correction at a public referral hospital in Northeast Brazil. In the postoperative period, the mean value of sTREM-1 was greater among patients diagnosed with sepsis than among those not diagnosed with sepsis (394.58 pg/mL versus 239.93 pg/mL, p < 0.001). Analysis of the ROC curve for sTREM-1 and sepsis revealed that the area under the curve was 0.761, with a 95% CI (0.587-0.935) and p = 0.013. With the Bayesian model, we found that a 100% probability of sepsis was related to postoperative blood concentrations of CRP above 71 mg/dL, a leukogram above 14,000 cells/μL, and sTREM-1 concentrations above the cutoff point (283.53 pg/mL). The proposed model using the Bayesian network approach with the combination of CRP, leukocyte count, and postoperative sTREM-1 showed promise for the diagnosis of sepsis.

A clinical decision support system (CDSS) is used as an aid in decision-making processes of health care providers in their day-to-day activities. This research attempts diagnosis of multiple diseases based on symptoms provided by patients. The work also recommends laboratory tests related to the predicted diseases and medications based on their results. The methodology adopted for implementation of CDSS is the Bayesian network approach. The modeling of the Bayesian network structure was undertaken in consultation with experts from medical domain. Clinical data has been used for estimation of network parameters such as conditional probability tables thereby bringing in machine learning into Bayesian methodology. The model developed is a learning model wherein the system input is saved for future training of the model. The results indicate that Bayesian approach is suitable for implementing a CDSS for multiple disease diagnosis. The proposed work will be useful towards increasing physicians throughput.

ABSTRACTBackground: Eye movement desensitization and reprocessing (EMDR) is an effective psychological treatment for posttraumatic stress disorder. Recalling a memory while simultaneously making eye movements (EM) decreases a memory’s vividness and/or emotionality. It has been argued that non-specific factors, such as treatment expectancy and experimental demand, may contribute to the EMDR’s effectiveness.Objective: The present study was designed to test whether expectations about the working mechanism of EMDR would alter the memory attenuating effects of EM. Two experiments were conducted. In Experiment 1, we examined the effects of pre-existing (non-manipulated) knowledge of EMDR in participants with and without prior knowledge. In Experiment 2, we experimentally manipulated prior knowledge by providing participants without prior knowledge with correct or incorrect information about EMDR’s working mechanism.Method: Participants in both experiments recalled two aversive, autobiographical memories during brief sets of EM (Recall+EM) or keeping eyes stationary (Recall Only). Before and after the intervention, participants scored their memories on vividness and emotionality. A Bayesian approach was used to compare two competing hypotheses on the effects of (existing/given) prior knowledge: (1) Prior (correct) knowledge increases the effects of Recall+EM vs. Recall Only, vs. (2) prior knowledge does not affect the effects of Recall+EM.Results: Recall+EM caused greater reductions in memory vividness and emotionality than Recall Only in all groups, including the incorrect information group. In Experiment 1, both hypotheses were supported by the data: prior knowledge boosted the effects of EM, but only modestly. In Experiment 2, the second hypothesis was clearly supported over the first: providing knowledge of the underlying mechanism of EMDR did not alter the effects of EM.Conclusions: Recall+EM appears to be quite robust against the effects of prior expectations. As Recall+EM is the core component of EMDR, expectancy effects probably contribute little to the effectiveness of EMDR treatment.

Discussion

EXP’s IC3D system classifies SAR or LiDAR derived 3D point clouds using a deep representation learning approach, producing as output a vector of categorical posterior probabilities of target classifications. Such posterior probabilities are suitable observational inputs to a Bayesian belief network (BBN), such as the EXP Shadow Compass system. In concert with conditional probabilities of intermediate events depending on the observation states, the Bayesian network computes posterior probabilities for events conditionally dependent on those intermediate event states. We demonstrate this approach by computing posterior event probabilities for sample analyst scenarios with intermediate event conditional probabilities specified by analysts. Future work could include extending the Bayesian network approach to discovery of the network topology from analyst data.