Papers by Keyword: Bayesian Network

Abstract: In Chile there are numerous drinking water projects in rural areas, from which it is possible to extract relevant information that could be incorporated into a learning model based on cases to estimate a particular variable. Among the learning models based on cases are Bayesian networks, which have the particularity to predict a variable and quantify its uncertainty. In this paper, a methodology to build a model based on Bayesian networks to estimate the likely investment cost of a new drinking water project in rural areas is proposed. It has a database of 32 projects built between 2009 and 2014, in the region of Los Rios, Chile. Two Bayesian networks structures were created, each with eight common variables to both. The proposed networks were trained with data extracted from 26 randomly selected projects. The remaining 6 projects were used as a control group to evaluate Bayesian networks and compare their results. When each network was evaluated with the control group, it was observed that in general the predicted results were consistent with those observed in 83% of cases. Finally, it was concluded that the constructed model can estimate the likely investment cost of a new drinking water project in rural areas, quantifying prediction uncertainty, expressing the results in probabilistic terms. This model could become a useful management tool, both companies and government agencies whose mission is to assess, allocate resources, and define projects approval and implementation.

Abstract: Based on the analysis of the advantages and limits of the traditional fault tree and Bayesian network in fault diagnosis, the method that building the fault Bayesian network based on fault tree is proposed in this paper. The paper introduces the correspondences between elements of the fault tree and the fault Bayesian network, also describes the inference process of the junction tree algorithm in the fault Bayesian network. Then with the foundation brake rigging system of CRH380AL EMU as an example, we build up the fault tree, complete its transmission to the fault Bayesian network, proving the superiority of the fault Bayesian tree in fault analysis of the complex system at last.

Abstract: The external traffic environment has a big influence to the traffic safety during the area of traffic conflict place，and to analysis the relationship between the external traffic environment factors and driving safety is helpful to improve the traffic safety. The method of comprehensive analysis the historical data and expert survey data is used to explore this question. And at the same time, the collision risk prediction model during the traffic conflict place is built by the Bayesian network. According to the data analyzing, the node variable, the state of variable and the conditional probability table of this model is also built. Finally, the software of Hugin is used to deal with the posteriori probability of collision risk, and the result proved that this model can predict the collision risk accurately during the traffic conflict area, and the data analyzing showed that the factor of the driver's intention, the vehicle speed and the headway have a significance influence to the traffic safety.

Abstract: In this thesis a remote intelligent fault diagnosis system based on wireless network is put forward. A method of data-transmission based on wireless network is analyzed and a scheme for intelligent fault diagnosis on the basis of Bayesian network is laid out.This thesis also researches on how to build a fault diagnosis model taking advantage of Bayesian network and how to improve the system's resolving ability and reasoning quality in the process of intelligent fault diagnosis.

Abstract: In order to improve damage diagnosis ability of maintenance personnel, constructing method of Bayesian network applied to weapon battlefield damage diagnosis is researched. Battlefield damage correlations among damaged parts of weapon are analyzed if one weapon is attacked by bombshells, and is the basis of damage diagnosis with the use of Bayesian network. Bayesian network for damage diagnosis is constructed based on K2 arithmetic. Variables sequence is the key factor of Bayesian network constructing, a statistical method of ascertaining variables sequence is presented with the use of weapon battlefield simulation technology.

Abstract: This paper develops a model validation method in the case of the full scale tests for the system model are infeasible. The Bayesian network with uncertain conditional probability parameters is used to represent the relations between the large computational model and its smaller modules. The interval probability theory is adopted to extrapolate the posterior probability of the interested variable in the uncertain Bayesian network. An interval valued Bayes factor is obtained to be the metric for model validation.

Abstract: In order to effectively analyze the mechanism for the occurrence of ship collision accidents caused by human factors, an accident causing chain was constructed using the Bayesian network structure and the data mining algorithm. According to navigator's cognitive behavior forming process and human errors, the accident cause network structure was constructed using the Bayesian network structure by analyzing 120 typical cases about ship collision accidents caused by human factors; a collision accident cause chain was obtained by mining the frequent combination of human errors using data-mining based Apriori algorithm and JAVA programming language.

Abstract: Cognitive maps represent decision makers’ mental maps and their strategies, which are always uncertain, ambiguous and hard to be formalized. In order to make intelligent design decision-making, a Bayesian approach for constructing cognitive maps is proposed in this paper. The cognitive map is modeled compatible with a Bayesian Network. Then cause-effect mapping rules between design elements embedded in cognitive maps can be made explicit by means of network structure learning. A score-based greedy search algorithm is implemented for network structure learning, in which penalized mutual information is defined as the scoring metric and hill-climbing search algorithm is used to find the highest-scoring network. The eliminating loop operator is introduced into the algorithm according to the restriction of the edge directionality.

Abstract: This paper describes an adaptive interface design that is suitable for process management system which is an important component of e-government. Bayesian network model is adopted for updating the layout of modules on the interface based on the user's operating data. Not only the number of modules will change, the positions of the modules can also update according to the calculation results. The experiment is conducted on a real college scientific research management system. Response time and click rates are used for evaluating the effectiveness of the algorithm, and the result shows that the value of these indicators declines obviously. Therefore, the method is proved to work.