Towards Tree-structure based Hierarchical Hybrid Soft Computing

s: There are three basic multilevel structures for hierarchical models, namely, incremental, aggregated and cascaded. Designing of these hierarchical models faces many difficulties including determination of the hierarchical structure, parameter identification and input variables selection for each submodels. A tree-structure based Hierarchical Hybrid Soft Computing (HHSC) framework is presented in this paper. Some instructions/operators for creating different HHSC models are discussed. An automatic design method of the HHSC models is proposed. In the framework, the tree-structure based evolutionary algorithms and a number of global search algorithms can be employed to optimize hierarchical structure, input selection and free parameters embedded in the HHSC model, respectively.