Research Of Auto Selection Of Model In Intelligent Decision Support System

IDSS is a burgeoning information system(IS), which artificial intelligence technology is imported in DSS. Model base system(MBS) and data base system(DBS) and knowledge base system(KBS) is foundation in IDSS. MBS is kernel part in DSS. MBS is made of model base and model base management system(MBMS). The major research contents have model controlling and model representation. Model selection acts as foundation of model controlling and plays an important role in MBMS.Emphasis of model selection exists in automatic selection of model and how to realize automatic selection is an important research domain in IDSS. Because model selection is a relatively difficult problem, most research achievement doesn't entirely realize automatic selection. Model selection mostly depended on expert and past selection experience. So research of model automatic selection has certain theory value and profound meaning in such situation. The problem about model automatic selection is divided in three parts, namely, selection of model type and selection of model structure and parameter ascertain.We build a classifier system by using the method of machine learning based genetic algorithms in selection of model type. The classifier differs from traditional classifier, which makes use of particle swarm confidence allocation algorithm. It has excellence such as algorithm simplicity and shortcut compared to traditional bucket queue algorithm. We adopts decision tree based on inductive inference methods in selection of model structure. To solve over-fitting problem in decision tree, we use a complex pruning method which pre-pruning combined with post-pruning in pruning algorithm of decision tree. We make use of genetic algorithm to confirm parameter with correspond to model structure in model instance ascertain.