Fusion of expert knowledge with data using belief functions: a case study in waste-water treatment

S. Populaire1, P. Ginestet2, J. Blanc1, T. Denoeux2
1Université de Technologie de Compiègne, UMR CNRS, Compiègne, France
2Ondeo Services Technical and Research Center, paris

Tóm tắt

This paper presents a methodology for combining expert knowledge with information from statistical data, in classification and prediction problems. The method is based on (1) a case-based approach allowing to predict a quantity of interest from past cases in the form of a belief function, (2) Bayesian networks for modelling expert knowledge and (3) a tuning mechanism allowing to optimally discount information sources by optimizing a performance criterion. This methodology is applied to the prediction of chemical oxygen demand solubility in waste-water The approach is expected to be useful in situations where both small databases and partial expert knowledge are available.

Từ khóa

#Computer aided software engineering #Wastewater treatment #Bayesian methods #Chemicals #Information resources #Predictive models #Optimization methods #Databases #Uncertainty

Tài liệu tham khảo

10.1109/5326.669565 vandegaag, 0, Probabilities for A Probabilistic Network A Case-study in Oesophageal Carcinoma 10.1016/0004-3702(90)90026-V shafer, 1976, A Mathematical Theory Ofevidence, 10.1515/9780691214696 10.1016/0004-3702(94)90026-4 pearl, 1988, Probabilistic Reasoning in Intelligent Systems Networks of Plausible Inference 10.1016/S0888-613X(99)00027-4 jensen, 1996, An Introduction to Bayesian Networks onisko, 2000, Learning bayesian networks parameters from small data sets: Application of noisy-or gates, Bayesian and Causal Networks From Inference to Data Mining' 12th European Conference on Artificial Intelligence 10.1016/S0165-0114(00)00086-5 10.1109/21.376493 charniak, 1991, Bayesian Networks Without Tears, 12, 50 henrion, 1987, Some practical issues in constructing belief networks, Uncertainty in Artificial Intelligence 3, 161 druzdzel, 0, Genie: A development environment for graphical decision-analytic models, Proceedings of the 1999Annual Symposium of the American Medical Informatics Association (AMIA-1999) 10.1109/TKDE.2000.868901 druzdzel, 1999, Knowledge engineering for very large decision-analytic medical models, Proceedings of the 1999 Annual Meeting of the American Medical Informatics Association, 10 druzdzel, 1996, Qualitative verbal explanations in bayesian belief networks, Al and Simulation of Behaviour Quarterly (Special Issue on Bayesian Belief Networks), 94, 43 elouedi, 2001, The evaluation of sensor's reliability and their tuning for multisensor data fusion within the transferable behef model, ECSQARU 2001, 305 druzdzel, 2000, Criteria for combining knowledge from different sources in probabilistic models, Working Notes of the UAI2000 Workshop on Domain Knowledge with Data for Decision Support