Bayesian Networks

Editors OLIVIER POURRET, Electricité de France PATRICK NAÏM, ELSEWARE, France BRUCE MARCOT, USDA Forest Service, Oregon, USA

• Foreword ixPreface xi1 Introduction to Bayesian networks 11.1 Models 11.2 Probabilistic vs. deterministic models 51.3 Unconditional and conditional independence 91.4 Bayesian networks 112 Medical diagnosis 152.1 Bayesian networks in medicine 152.2 Context and history 172.3 Model construction 192.4 Inference 262.5 Model validation 282.6 Model use 302.7 Comparison to other approaches 312.8 Conclusions and perspectives 323 Clinical decision support 333.1 Introduction 333.2 Models and methodology 343.3 The Busselton network 353.4 The PROCAM network 403.5 The PROCAM Busselton network 443.6 Evaluation 463.7 The clinical support tool: TakeHeartII 473.8 Conclusion 514 Complex genetic models 534.1 Introduction 534.2 Historical perspectives 544.3 Complex traits 564.4 Bayesian networks to dissect complex traits 594.5 Applications 644.6 Future challenges 715 Crime risk factors analysis 735.1 Introduction 735.2 Analysis of the factors affecting crime risk 745.3 Expert probabilities elicitation 755.4 Data preprocessing 765.5 A Bayesian network model 785.6 Results 805.7 Accuracy assessment 835.8 Conclusions 846 Spatial dynamics in France 876.1 Introduction 876.2 An indicator-based analysis 896.3 The Bayesian network model 976.4 Conclusions 1097 Inference problems in forensic science 1137.1 Introduction 1137.2 Building Bayesian networks for inference 1167.3 Applications of Bayesian networks in forensic science 1207.4 Conclusions 1268 Conservation of marbled murrelets in British Columbia 1278.1 Context/history 1278.2 Model construction 1298.3 Model calibration, validation and use 1368.4 Conclusions/perspectives 1479 Classifiers for modeling of mineral potential 1499.1 Mineral potential mapping 1499.2 Classifiers for mineral potential mapping 1519.3 Bayesian network mapping of base metal deposit 1579.4 Discussion 1669.5 Conclusions 17110 Student modeling 17310.1 Introduction 17310.2 Probabilistic relational models 17510.3 Probabilistic relational student model 17610.4 Case study 18010.5 Experimental evaluation 18210.6 Conclusions and future directions 18511 Sensor validation 18711.1 Introduction 18711.2 The problem of sensor validation 18811.3 Sensor validation algorithm 19111.4 Gas turbines 19711.5 Models learned and experimentation 19811.6 Discussion and conclusion 20212 An information retrieval system 20312.1 Introduction 20312.2 Overview 20512.3 Bayesian networks and information retrieval 20612.4 Theoretical foundations 20712.5 Building the information retrieval system 21512.6 Conclusion 22313 Reliability analysis of systems 22513.1 Introduction 22513.2 Dynamic fault trees 22713.3 Dynamic Bayesian networks 22813.4 A case study: The Hypothetical Sprinkler System 23013.5 Conclusions 23714 Terrorism risk management 23914.1 Introduction 24014.2 The Risk Influence Network 25014.3 Software implementation 25414.4 Site Profiler deployment 25914.5 Conclusion 26115 Credit-rating of companies 26315.1 Introduction 26315.2 Naive Bayesian classifiers 26415.3 Example of actual credit-ratings systems 26415.4 Credit-rating data of Japanese companies 26615.5 Numerical experiments 26715.6 Performance comparison of classifiers 27315.7 Conclusion 27616 Classification of Chilean wines 27916.1 Introduction 27916.2 Experimental setup 28116.3 Feature extraction methods 28516.4 Classification results 28816.5 Conclusions 29817 Pavement and bridge management 30117.1 Introduction 30117.2 Pavement management decisions 30217.3 Bridge management 30717.4 Bridge approach embankment – case study 30817.5 Conclusion 31218 Complex industrial process operation 31318.1 Introduction 31318.2 A methodology for Root Cause Analysis 31418.3 Pulp and paper application 32118.4 The ABB Industrial IT platform 32518.5 Conclusion 32619 Probability of default for large corporates 32919.1 Introduction 32919.2 Model construction 33219.3 BayesCredit 33519.4 Model benchmarking 34119.5 Benefits from technology and software 34219.6 Conclusion 34320 Risk management in robotics 34520.1 Introduction 34520.2 DeepC 34620.3 The ADVOCATE II architecture 35220.4 Model development 35420.5 Model usage and examples 36020.6 Benefits from using probabilistic graphical models 36120.7 Conclusion 36221 Enhancing Human Cognition 36521.1 Introduction 36521.2 Human foreknowledge in everyday settings 36621.3 Machine foreknowledge 36921.4 Current application and future research needs 37321.5 Conclusion 37522 Conclusion 37722.1 An artificial intelligence perspective 37722.2 A rational approach of knowledge 37922.3 Future challenges 384Bibliography 385Index 427