Design of hybrid differential evolution and group method of data handling networks for modeling and prediction

摘要: This paper proposes a hybrid modeling approach based on two familiar non-linear methods of mathematical modeling; the group method data handling (GMDH) and differential evolution (DE) population-based algorithm. The proposed constructs GMDH self-organizing network model population promising DE solutions. new implementation is then applied to tool wear in milling operations also representative time series prediction problems exchange rates three international currencies well-studied Box-Jenkins gas furnace process data. results DE-GMDH are compared with obtained by standard algorithm its variants. Results presented show that appears perform better than polynomial neural (PNN) for problem. For rate problem, competitive all other approaches except one case. data, experimental clearly demonstrates DE-GMDH-type outperforms existing models both terms approximation capabilities as well generalization abilities. Consequently, this may be useful advanced manufacturing systems where it necessary during machining operations, applications such industrial problems.