Research of Model Structure Selection Based on Genetic Algorithm

Model management is the key of the success and practicality of decision-making support system. As the base of model management, model selection plays an important role in model management. In model selection, automatic binary decision tree construction based on genetic algorithm for model selection is proposed. In binary decision tree, each terminal node corresponds to a specific data type and represents a specific model. The decision function of non-terminal node is used to decide path until the model structure is determined. The method has been successfully adopted to quality defect diagnosis warning model management of a steel company.