Model Generation, Verification and Revision

Next: System Configuration Up: Framework Previous: Environmental System Modeling

Model Generation, Verification and Revision

We propose a framework designed to support the process of deriving a model for each cell. The modeling process consists of three major phases:

Model Hypothesis: During this phase, domain experts identify potential factors for the spread of the disease. Factors considered for Hantavirus Pulmonary Syndrome (HPS) include: outbreaks of the disease in neighboring areas, the prevalence of suburban houses that are surrounded by bushes, and specific climate patterns such as a wet summer followed by a dry summer. Although factors are identified for the model, the precise impacts (or the associated weights) are not yet determined during this phase. The hypothetical model, which includes both a localized model (Fig. 2) and a global model (Fig. 1) is captured through a set of composite object queries to be described in Section 4.
Model Verification/Validation: During this phase, the relevance of each individual factor is quantitatively identified. The complete model is then verified by applying it to historical data and comparing the prediction past occurrences of the disease. The model prediction is carried through executing the composite object queries formulated during the model generation phase, and will be described in detail in Section 5. The verification results are used to revise the original model.
Model Revision: After the model verification, adjustments are made to bring the predicted results more in line with the validation data set. If the revision is insignificant, then the model has already converged, and the revision process ends. Otherwise, another round of model verification/validation is initiated. This will be described in more detail in Section 6.