Friday, April 3, 2009

Process of Model Evaluation

Steps in Model Evaluation
  1. Verification: To ensure that the model behave as the experimenter intends
  2. Validation: To test the agreement between the behavior of the model and real system. Most difficult and time-consuming. Mistakenly waited until model is complete.
  3. Problem Analysis: Deals with analysis and interpretation of the data generated by the experiments
Commonly Used Validation Methods:
  • Face Value Approach: Method of approving rules (prior to their implementation in models) through face-value. A panel of experts make an outline of how the rules should be developed and most agreed upon rules are written in a rule book that serves as a means of validation.
  • Model Vs Real World Comparison Approach: Compare outputs of real world and the model using identical inputs. This process is never possible but attempted to some degree.
Some Guidelines:
  • Continual verification and validation are recommended.
  • Analysis of input & output data and V&V of models, all are equally important

No comments: