Given a random sample from a multivariate probability distribution p, the maximum minimum parents and children algorithm locates the skeleton of the directed acyclic graph of a Bayesian network for p provided that there exists a faithful Bayesian network and that the dependence structure derived from data is the same as that of the underlying probability distribution. The aim of this thesis is to examine the consequences when one of these conditions is not fulfilled. There are some circumstances where the algorithm works well even if there does not exist a faithful Bayesian network, but there are others where the algorithm fails. The MMPC tests for conditional independence between the variables and assumes that if conditional independence is not rejected, then the conditional independence statement holds. There are situations where this procedure leads to conditional independence being accepted that contradict conditional dependence relations in the data. This leads to edges being removed from the skeleton that are necessary for representing the dependence structure of the data.

In this doctoral thesis, I present my research into applying machine learning techniques for reconstructing species interaction networks in ecology, reconstructing molecular signalling pathways and gene regulatory networks in systems biology, and inferring parameters in ordinary differential equation (ODE) models of signalling pathways. Together, the methods I have developed for these applications demonstrate the usefulness of machine learning for reconstructing networks and inferring network parameters from data. The thesis consists of three parts. The first part is a detailed comparison of applying static Bayesian networks, relevance vector machines, and linear regression with L1 regularisation (LASSO) to the problem of reconstructing species interaction networks from species absence/presence data in ecology (Faisal et al., 2010). I describe how I generated data from a stochastic population model to test the different methods and how the simulation study led us to introduce spatial autocorrelation as an important covariate. I also show how we used the results of the simulation study to apply the methods to presence/absence data of bird species from the European Bird Atlas. The second part of the thesis describes a time-varying, non-homogeneous dynamic Bayesian network model for reconstructing signalling pathways and gene regulatory networks, based on L`ebre et al. (2010). I show how my work has extended this model to incorporate different types of hierarchical Bayesian information sharing priors and different coupling strategies among nodes in the network. The introduction of these priors reduces the inference uncertainty by putting a penalty on the number of structure changes among network segments separated by inferred changepoints (Dondelinger et al., 2010; Husmeier et al., 2010; Dondelinger et al., 2012b). Using both synthetic and real data, I demonstrate that using information sharing priors leads to a better reconstruction accuracy of the underlying gene regulatory networks, and I compare the different priors and coupling strategies. I show the results of applying the model to gene expression datasets from Drosophila melanogaster and Arabidopsis thaliana, as well as to a synthetic biology gene expression dataset from Saccharomyces cerevisiae. In each case, the underlying network is time-varying; for Drosophila melanogaster, as a consequence of measuring gene expression during different developmental stages; for Arabidopsis thaliana, as a consequence of measuring gene expression for circadian clock genes under different conditions; and for the synthetic biology dataset, as a consequence of changing the growth environment. I show that in addition to inferring sensible network structures, the model also successfully predicts the locations of changepoints. The third and final part of this thesis is concerned with parameter inference in ODE models of biological systems. This problem is of interest to systems biology researchers, as kinetic reaction parameters can often not be measured, or can only be estimated imprecisely from experimental data. Due to the cost of numerically solving the ODE system after each parameter adaptation, this is a computationally challenging problem. Gradient matching techniques circumvent this problem by directly fitting the derivatives of the ODE to the slope of an interpolant. I present an inference procedure for a model using nonparametric Bayesian statistics with Gaussian processes, based on Calderhead et al. (2008). I show that the new inference procedure improves on the original formulation in Calderhead et al. (2008) and I present the result of applying it to ODE models of predator-prey interactions, a circadian clock gene, a signal transduction pathway, and the JAK/STAT pathway.

Individualmente, as linhas de pesquisa da Inteligência Artificial têm proposto abordagens para a resolução de inúmeros problemas complexos do mundo real. O paradigma orientado a agentes provê os agentes autônomos, capazes de perceber os seus ambientes, reagir de acordo com diferentes circunstâncias e estabelecer interações sociais com outros agentes de software ou humanos. As redes Bayesianas fornecem uma maneira de representar graficamente as distribuições de probabilidades condicionais e permitem a realização de raciocínios probabilísticos baseados em evidências. As ontologias são especificações explícitas e formais de conceituações, que são usadas em uma variedade de áreas de pesquisa, incluindo os Sistemas Multiagentes. Contudo, existem aplicações cujos requisitos não podem ser atendidos por uma única tecnologia. Circunstâncias como estas exigem a integração de tecnologias desenvolvidas por distintas áreas da Ciência da Computação. Esta dissertação trata a integração do modelo de agentes BDI (Belief-Desire-Intention) e das redes Bayesianas. Além disso, é adotada uma abordagem baseada em ontologias para representar o conhecimento incerto dos agentes. O primeiro passo em direção a integração foi o desenvolvimento de uma ontologia para representar a estrutura das redes Bayesinas. Esta ontologia tem como principal objetivo permitir a interoperabilidade agentes compatíveis com a arquitetura proposta. No entanto, a ontologia também facilita o entendimento necessário para abstrair os estados mentais e processos cognitivos dos agentes através de elementos das redes Bayesianas. Uma vez construída a ontologia, a mesma foi integrada com a arquitetura BDI. Através da integração do modelo BDI com as redes Bayesianas foi obtida uma arquitetura cognitiva de agentes capaz de deliberar sob incerteza. O processo de integração foi composto de duas etapas: abstração dos estados mentais através de elementos das redes Bayesianas e especificação do processo deliberativo. Finalmente, foi desenvolvido um estudo de caso, que consistiu na aplicação da arquitetura proposta no Agente Social, um componente de um portal educacional multiagente (PortEdu). / Individually, Artificial Intelligence research areas have proposed approaches to solve several complex real-world problems. The agent-based paradigm provided autonomous agents, capable of perceiving their environment, reacting in accordance with different situations, and establishing social interactions with other software agents and humans. Bayesian networks provided a way to represent graphically the conditional probability distributions and an evidence-based probabilistic reasoning. Ontologies are an effort to develop formal and explicit specifications of concepts, which have been used by a wide range of research areas, including Multiagent Systems. However, there are applications whose requirements can not be addressed by a single technology. Circumstances like these demand the integration of technologies developed by distinct areas of Computer Science. This work is particularly concerned with the integration of Belief-Desire-Intention (BDI) agent architecture and Bayesian networks. Moreover, it is adopted an ontology-based approach to represent the agent’s uncertain knowledge. To bring together those technologies, it was developed an ontology to represent the structure of Bayesian networks knowledge representation. This ontology supports the interoperability among agents that comply with the proposed architecture, and it also facilitates the understanding necessary to abstract the agents’ mental states and cognitive processes through elements of Bayesian networks. Once specified the ontology, it was integrated with the BDI agent architecture. By integrating BDI architecture and Bayesian networks, it was obtained a cognitive agent architecture capable of reasoning under uncertainty. It was performed in two stages: abstraction of mental states through Bayesian networks and specification of the deliberative process. Finally, it was developed a case study, which consists in applying the probabilistic BDI architecture in the Social Agent, a component of a multiagent educational portal (PortEdu).

O presente trabalho emprega Redes Bayesianas para a previsão da rentabilidade de empresas. Define-se como rentabilidade superior as empresa que obtiveram retorno para os acionistas classificados acima de 81,5% em relação às demais. Adota-se a metodologia de seleção dos indicadores proposta por Sun e Shenoy (2007), que seleciona as variáveis explicativas segundo suas correlações com a variável classificadora. Obtêm-se, ao final, dois modelos sendo o primeiro com dois estados de classificação de empresas, superior e inferior; o segundo com três estados (superior mediano e inferior). Assim como Sun e Shenoy (2007), tenta-se validar o modelo Bayesiano com a regressão logística. Constata-se que não é possível afirmar que as média das taxas de sucesso dos dois modelos sejam diferentes ao se prever rentabilidade superior, entretanto a regressão tem melhor desempenho ao se prever rentabilidade baixa. A variável mais significativa tanto para o primeiro quanto para o segundo modelos foi a classificação atual da empresa, ou seja, empresas que figuram em um determinado ano no estado de rentabilidade superior são as mais propensas a repetir o resultado do que as demais. Os resultados apontam taxas de acerto que vão de 14,70% em 1999 (ano da crise cambial quando a rentabilidade média das empresas foi de 2,74%) a 52,94% em 1997 (ano cuja rentabilidade média foi de 11,76%) para o primeiro modelo e de 11,76 % (1999) a 56,60 % (2004, rentabilidade média de 10,76%) para o segundo modelo. Apesar dos modelos ainda não conseguirem alcançar uma estabilidade nas previsões os resultados são animadores quando se desenvolve a hipótese de utilidade para um possível investidor e a expectativa de retorno acumulado, ao longo dos dez anos, passa de 70,37%, que é a rentabilidade média acumulada do período, para 357,07% e 410,10 % para o primeiro e o segundo modelo respectivamente. / This work use the knowledge obtained from Bayesian networks studies of bankruptcy prediction and applied it for forecasting companies' profitability. Higher profitability is defined as the company that had returns for shareholders classified over 81.5% compared to the others. Adopting the methodology of selection of the explanatory variables proposed by Sun and SHENOY (2007) based on correlations among them with the classification variable. As a result it is obtained two models, the first one with two classification states for de classification variable, upper and low, and the second one with three states (upper, middle and low). As Sun and SHENOY (2007), the Bayesian model was compared with a logistic regression. It cannot be say that the average success rates of the two models are different for forecasting higher profitability; otherwise, for low profitability forecasts the regression model was superior. The most significant variable for both the first and for the second model was the previous company's return for the shareholders, i.e. companies that are in a given year in the state of upper profitability are more likely to repeat the resulting the next year. The results show success rates ranging from 14.70% in 1999 (year of the currency crisis when the average profitability of the companies was 2.74%) to 52.94% in 1997 (average return rate was 11.76 %) for the first model and from 11.76% (1999) to 56.60% (2004, average return rate was 10.76%) for the second model. Although the models still fail to achieve stability in the estimates the results are encouraging when developing the hypothesis of possible investor profitability when the expectation of return accumulated over the ten years, range from 70.37%, which is the average profitability accumulated in the period to 357.07% and 410.10% respectively for the first and second model.

We describe an online clinical decision support (CDS) system, Lung Cancer Assistant (LCA), which we have developed to aid the clinicians in arriving at informed treatment decisions for lung cancer patients at multidisciplinary team (MDT) meetings. LCA integrates rule-based and probabilistic decision support within a single platform. To our knowledge, this is the first time this has been achieved in the context of CDS in cancer care. Rule-based decision support is achieved by an original ontological guideline rule inference framework that operates on a domain-specific module of Systematized Nomenclature of Medicine-Clinical Terms (SNOMED-CT), containing clinical concepts and guideline rule knowledge elicited from the major national and international guideline publishers. It adopts a conventional argumentation-based decision model, whereby the decision options are listed along with arguments derived by matching the patient records to the guideline rule base. As an additional feature of this framework, when a new patient is entered, LCA displays the most similar patients to the one being viewed. Probabilistic inference is provided by a Bayesian Network (BN) whose structure and parameters have been learned based on the English Lung Cancer Database (LUCADA). This allows LCA to predict the probability of patient survival and lay out how the selection of different treatment plans would affect it. Based on a retrospective patient subset from LUCADA, we present empirical results on the treatment recommendations provided by both functionalities of LCA and discuss their strengths and weaknesses. Finally, we present preliminary work, which may allow utilising the BN to calculate survival odd ratios that could be translated into quantitative degrees of support for the guideline rule-based arguments. An online version of LCA is accessible on http://lca.eng.ox.ac.uk.

O presente trabalho emprega Redes Bayesianas para a previsão da rentabilidade de empresas. Define-se como rentabilidade superior as empresa que obtiveram retorno para os acionistas classificados acima de 81,5% em relação às demais. Adota-se a metodologia de seleção dos indicadores proposta por Sun e Shenoy (2007), que seleciona as variáveis explicativas segundo suas correlações com a variável classificadora. Obtêm-se, ao final, dois modelos sendo o primeiro com dois estados de classificação de empresas, superior e inferior; o segundo com três estados (superior mediano e inferior). Assim como Sun e Shenoy (2007), tenta-se validar o modelo Bayesiano com a regressão logística. Constata-se que não é possível afirmar que as média das taxas de sucesso dos dois modelos sejam diferentes ao se prever rentabilidade superior, entretanto a regressão tem melhor desempenho ao se prever rentabilidade baixa. A variável mais significativa tanto para o primeiro quanto para o segundo modelos foi a classificação atual da empresa, ou seja, empresas que figuram em um determinado ano no estado de rentabilidade superior são as mais propensas a repetir o resultado do que as demais. Os resultados apontam taxas de acerto que vão de 14,70% em 1999 (ano da crise cambial quando a rentabilidade média das empresas foi de 2,74%) a 52,94% em 1997 (ano cuja rentabilidade média foi de 11,76%) para o primeiro modelo e de 11,76 % (1999) a 56,60 % (2004, rentabilidade média de 10,76%) para o segundo modelo. Apesar dos modelos ainda não conseguirem alcançar uma estabilidade nas previsões os resultados são animadores quando se desenvolve a hipótese de utilidade para um possível investidor e a expectativa de retorno acumulado, ao longo dos dez anos, passa de 70,37%, que é a rentabilidade média acumulada do período, para 357,07% e 410,10 % para o primeiro e o segundo modelo respectivamente. / This work use the knowledge obtained from Bayesian networks studies of bankruptcy prediction and applied it for forecasting companies' profitability. Higher profitability is defined as the company that had returns for shareholders classified over 81.5% compared to the others. Adopting the methodology of selection of the explanatory variables proposed by Sun and SHENOY (2007) based on correlations among them with the classification variable. As a result it is obtained two models, the first one with two classification states for de classification variable, upper and low, and the second one with three states (upper, middle and low). As Sun and SHENOY (2007), the Bayesian model was compared with a logistic regression. It cannot be say that the average success rates of the two models are different for forecasting higher profitability; otherwise, for low profitability forecasts the regression model was superior. The most significant variable for both the first and for the second model was the previous company's return for the shareholders, i.e. companies that are in a given year in the state of upper profitability are more likely to repeat the resulting the next year. The results show success rates ranging from 14.70% in 1999 (year of the currency crisis when the average profitability of the companies was 2.74%) to 52.94% in 1997 (average return rate was 11.76 %) for the first model and from 11.76% (1999) to 56.60% (2004, average return rate was 10.76%) for the second model. Although the models still fail to achieve stability in the estimates the results are encouraging when developing the hypothesis of possible investor profitability when the expectation of return accumulated over the ten years, range from 70.37%, which is the average profitability accumulated in the period to 357.07% and 410.10% respectively for the first and second model.

abstract: Bayesian networks are powerful tools in system reliability assessment due to their flexibility in modeling the reliability structure of complex systems. This dissertation develops Bayesian network models for system reliability analysis through the use of Bayesian inference techniques. Bayesian networks generalize fault trees by allowing components and subsystems to be related by conditional probabilities instead of deterministic relationships; thus, they provide analytical advantages to the situation when the failure structure is not well understood, especially during the product design stage. In order to tackle this problem, one needs to utilize auxiliary information such as the reliability information from similar products and domain expertise. For this purpose, a Bayesian network approach is proposed to incorporate data from functional analysis and parent products. The functions with low reliability and their impact on other functions in the network are identified, so that design changes can be suggested for system reliability improvement. A complex system does not necessarily have all components being monitored at the same time, causing another challenge in the reliability assessment problem. Sometimes there are a limited number of sensors deployed in the system to monitor the states of some components or subsystems, but not all of them. Data simultaneously collected from multiple sensors on the same system are analyzed using a Bayesian network approach, and the conditional probabilities of the network are estimated by combining failure information and expert opinions at both system and component levels. Several data scenarios with discrete, continuous and hybrid data (both discrete and continuous data) are analyzed. Posterior distributions of the reliability parameters of the system and components are assessed using simultaneous data. Finally, a Bayesian framework is proposed to incorporate different sources of prior information and reconcile these different sources, including expert opinions and component information, in order to form a prior distribution for the system. Incorporating expert opinion in the form of pseudo-observations substantially simplifies statistical modeling, as opposed to the pooling techniques and supra Bayesian methods used for combining prior distributions in the literature. The methods proposed are demonstrated with several case studies. / Dissertation/Thesis / Doctoral Dissertation Industrial Engineering 2016

Individualmente, as linhas de pesquisa da Inteligência Artificial têm proposto abordagens para a resolução de inúmeros problemas complexos do mundo real. O paradigma orientado a agentes provê os agentes autônomos, capazes de perceber os seus ambientes, reagir de acordo com diferentes circunstâncias e estabelecer interações sociais com outros agentes de software ou humanos. As redes Bayesianas fornecem uma maneira de representar graficamente as distribuições de probabilidades condicionais e permitem a realização de raciocínios probabilísticos baseados em evidências. As ontologias são especificações explícitas e formais de conceituações, que são usadas em uma variedade de áreas de pesquisa, incluindo os Sistemas Multiagentes. Contudo, existem aplicações cujos requisitos não podem ser atendidos por uma única tecnologia. Circunstâncias como estas exigem a integração de tecnologias desenvolvidas por distintas áreas da Ciência da Computação. Esta dissertação trata a integração do modelo de agentes BDI (Belief-Desire-Intention) e das redes Bayesianas. Além disso, é adotada uma abordagem baseada em ontologias para representar o conhecimento incerto dos agentes. O primeiro passo em direção a integração foi o desenvolvimento de uma ontologia para representar a estrutura das redes Bayesinas. Esta ontologia tem como principal objetivo permitir a interoperabilidade agentes compatíveis com a arquitetura proposta. No entanto, a ontologia também facilita o entendimento necessário para abstrair os estados mentais e processos cognitivos dos agentes através de elementos das redes Bayesianas. Uma vez construída a ontologia, a mesma foi integrada com a arquitetura BDI. Através da integração do modelo BDI com as redes Bayesianas foi obtida uma arquitetura cognitiva de agentes capaz de deliberar sob incerteza. O processo de integração foi composto de duas etapas: abstração dos estados mentais através de elementos das redes Bayesianas e especificação do processo deliberativo. Finalmente, foi desenvolvido um estudo de caso, que consistiu na aplicação da arquitetura proposta no Agente Social, um componente de um portal educacional multiagente (PortEdu). / Individually, Artificial Intelligence research areas have proposed approaches to solve several complex real-world problems. The agent-based paradigm provided autonomous agents, capable of perceiving their environment, reacting in accordance with different situations, and establishing social interactions with other software agents and humans. Bayesian networks provided a way to represent graphically the conditional probability distributions and an evidence-based probabilistic reasoning. Ontologies are an effort to develop formal and explicit specifications of concepts, which have been used by a wide range of research areas, including Multiagent Systems. However, there are applications whose requirements can not be addressed by a single technology. Circumstances like these demand the integration of technologies developed by distinct areas of Computer Science. This work is particularly concerned with the integration of Belief-Desire-Intention (BDI) agent architecture and Bayesian networks. Moreover, it is adopted an ontology-based approach to represent the agent’s uncertain knowledge. To bring together those technologies, it was developed an ontology to represent the structure of Bayesian networks knowledge representation. This ontology supports the interoperability among agents that comply with the proposed architecture, and it also facilitates the understanding necessary to abstract the agents’ mental states and cognitive processes through elements of Bayesian networks. Once specified the ontology, it was integrated with the BDI agent architecture. By integrating BDI architecture and Bayesian networks, it was obtained a cognitive agent architecture capable of reasoning under uncertainty. It was performed in two stages: abstraction of mental states through Bayesian networks and specification of the deliberative process. Finally, it was developed a case study, which consists in applying the probabilistic BDI architecture in the Social Agent, a component of a multiagent educational portal (PortEdu).

Individualmente, as linhas de pesquisa da Inteligência Artificial têm proposto abordagens para a resolução de inúmeros problemas complexos do mundo real. O paradigma orientado a agentes provê os agentes autônomos, capazes de perceber os seus ambientes, reagir de acordo com diferentes circunstâncias e estabelecer interações sociais com outros agentes de software ou humanos. As redes Bayesianas fornecem uma maneira de representar graficamente as distribuições de probabilidades condicionais e permitem a realização de raciocínios probabilísticos baseados em evidências. As ontologias são especificações explícitas e formais de conceituações, que são usadas em uma variedade de áreas de pesquisa, incluindo os Sistemas Multiagentes. Contudo, existem aplicações cujos requisitos não podem ser atendidos por uma única tecnologia. Circunstâncias como estas exigem a integração de tecnologias desenvolvidas por distintas áreas da Ciência da Computação. Esta dissertação trata a integração do modelo de agentes BDI (Belief-Desire-Intention) e das redes Bayesianas. Além disso, é adotada uma abordagem baseada em ontologias para representar o conhecimento incerto dos agentes. O primeiro passo em direção a integração foi o desenvolvimento de uma ontologia para representar a estrutura das redes Bayesinas. Esta ontologia tem como principal objetivo permitir a interoperabilidade agentes compatíveis com a arquitetura proposta. No entanto, a ontologia também facilita o entendimento necessário para abstrair os estados mentais e processos cognitivos dos agentes através de elementos das redes Bayesianas. Uma vez construída a ontologia, a mesma foi integrada com a arquitetura BDI. Através da integração do modelo BDI com as redes Bayesianas foi obtida uma arquitetura cognitiva de agentes capaz de deliberar sob incerteza. O processo de integração foi composto de duas etapas: abstração dos estados mentais através de elementos das redes Bayesianas e especificação do processo deliberativo. Finalmente, foi desenvolvido um estudo de caso, que consistiu na aplicação da arquitetura proposta no Agente Social, um componente de um portal educacional multiagente (PortEdu). / Individually, Artificial Intelligence research areas have proposed approaches to solve several complex real-world problems. The agent-based paradigm provided autonomous agents, capable of perceiving their environment, reacting in accordance with different situations, and establishing social interactions with other software agents and humans. Bayesian networks provided a way to represent graphically the conditional probability distributions and an evidence-based probabilistic reasoning. Ontologies are an effort to develop formal and explicit specifications of concepts, which have been used by a wide range of research areas, including Multiagent Systems. However, there are applications whose requirements can not be addressed by a single technology. Circumstances like these demand the integration of technologies developed by distinct areas of Computer Science. This work is particularly concerned with the integration of Belief-Desire-Intention (BDI) agent architecture and Bayesian networks. Moreover, it is adopted an ontology-based approach to represent the agent’s uncertain knowledge. To bring together those technologies, it was developed an ontology to represent the structure of Bayesian networks knowledge representation. This ontology supports the interoperability among agents that comply with the proposed architecture, and it also facilitates the understanding necessary to abstract the agents’ mental states and cognitive processes through elements of Bayesian networks. Once specified the ontology, it was integrated with the BDI agent architecture. By integrating BDI architecture and Bayesian networks, it was obtained a cognitive agent architecture capable of reasoning under uncertainty. It was performed in two stages: abstraction of mental states through Bayesian networks and specification of the deliberative process. Finally, it was developed a case study, which consists in applying the probabilistic BDI architecture in the Social Agent, a component of a multiagent educational portal (PortEdu).

O presente trabalho emprega Redes Bayesianas para a previsão da rentabilidade de empresas. Define-se como rentabilidade superior as empresa que obtiveram retorno para os acionistas classificados acima de 81,5% em relação às demais. Adota-se a metodologia de seleção dos indicadores proposta por Sun e Shenoy (2007), que seleciona as variáveis explicativas segundo suas correlações com a variável classificadora. Obtêm-se, ao final, dois modelos sendo o primeiro com dois estados de classificação de empresas, superior e inferior; o segundo com três estados (superior mediano e inferior). Assim como Sun e Shenoy (2007), tenta-se validar o modelo Bayesiano com a regressão logística. Constata-se que não é possível afirmar que as média das taxas de sucesso dos dois modelos sejam diferentes ao se prever rentabilidade superior, entretanto a regressão tem melhor desempenho ao se prever rentabilidade baixa. A variável mais significativa tanto para o primeiro quanto para o segundo modelos foi a classificação atual da empresa, ou seja, empresas que figuram em um determinado ano no estado de rentabilidade superior são as mais propensas a repetir o resultado do que as demais. Os resultados apontam taxas de acerto que vão de 14,70% em 1999 (ano da crise cambial quando a rentabilidade média das empresas foi de 2,74%) a 52,94% em 1997 (ano cuja rentabilidade média foi de 11,76%) para o primeiro modelo e de 11,76 % (1999) a 56,60 % (2004, rentabilidade média de 10,76%) para o segundo modelo. Apesar dos modelos ainda não conseguirem alcançar uma estabilidade nas previsões os resultados são animadores quando se desenvolve a hipótese de utilidade para um possível investidor e a expectativa de retorno acumulado, ao longo dos dez anos, passa de 70,37%, que é a rentabilidade média acumulada do período, para 357,07% e 410,10 % para o primeiro e o segundo modelo respectivamente. / This work use the knowledge obtained from Bayesian networks studies of bankruptcy prediction and applied it for forecasting companies' profitability. Higher profitability is defined as the company that had returns for shareholders classified over 81.5% compared to the others. Adopting the methodology of selection of the explanatory variables proposed by Sun and SHENOY (2007) based on correlations among them with the classification variable. As a result it is obtained two models, the first one with two classification states for de classification variable, upper and low, and the second one with three states (upper, middle and low). As Sun and SHENOY (2007), the Bayesian model was compared with a logistic regression. It cannot be say that the average success rates of the two models are different for forecasting higher profitability; otherwise, for low profitability forecasts the regression model was superior. The most significant variable for both the first and for the second model was the previous company's return for the shareholders, i.e. companies that are in a given year in the state of upper profitability are more likely to repeat the resulting the next year. The results show success rates ranging from 14.70% in 1999 (year of the currency crisis when the average profitability of the companies was 2.74%) to 52.94% in 1997 (average return rate was 11.76 %) for the first model and from 11.76% (1999) to 56.60% (2004, average return rate was 10.76%) for the second model. Although the models still fail to achieve stability in the estimates the results are encouraging when developing the hypothesis of possible investor profitability when the expectation of return accumulated over the ten years, range from 70.37%, which is the average profitability accumulated in the period to 357.07% and 410.10% respectively for the first and second model.