In this paper, we propose a new modeling strategy to deal with a complex system which has various sources of input data sets. Through a thorough study of existing aggregating methods of models, we found that most works only consider aggregation of local models with the same level of granularity. With this motivation, we construct a granular hierarchy model by local models with different levels of granularities. The constructed hierarchy model includes two levels: a lower level and a higher level. The global model in the higher level is aggregated with local models and information granularities. In this case, more abstract data (information) are used as the input of global model which assures the general output of the whole model. Knowledge reconciliation and interaction are crucial issues during the whole process. Hence we propose a method by analyzing prototypes of local models with the principle of justifiable granularity.
展开▼
