IDENTIFICATION OF THE IMPORTANT PARAMETERS OF COMPLEX SYSTEMS BY A BAYESIAN APPROACH
This work presents a method of variable selection based on the application of the Bayes theorem in order to take into account all the a priori information available associated with the engineering knowledge. In particular, the importance of each variable of the model, e.g. a regression model, is expressed by assigning to the corresponding system parameter a probability density function whose mean and variance are a priori established according to the judgement of experts. The value of the variance takes into account the uncertainty connected with the subjectivity of the judgement. Such information is introduced in the model and, then, the Bayes theorem applied after an expressly designed observation campaign. The parameters of the resulting a posteriori density functions provide updated values of the rank order of the variables and of the uncertainty associated with this evaluation.
Bibliographic Reference: 5TH EUREDATA CONFERENCE, HEIDELBERG (GERMANY), APRIL 9-11, 1986 WRITE TO CEC LUXEMBOURG, DG XIII/A2, POB 1907 MENTIONING PAPER E 32313 ORA
Availability: Can be ordered online
Record Number: 1989125016300 / Last updated on: 1987-02-01
Available languages: en