Research On Progressive Modeling Method Of Hierarchical Fuzzy Inference System Based On Data

Hierarchical fuzzy inference system has been widely used in modeling,control,decision making and other fields.However,the problem of high dimension is often encountered in practical applications.In the face of high dimension problems,progressive constructing hierarchical fuzzy inference system will make the modeling process more reasonable.Progressively construct fuzzy sub-system of each layer and added into the target fuzzy inference system.When the target system reaches the accuracy requirement,the construction will be stopped.In the above process,it is necessary to consider the input variables of sub-system of each layer and the problem of fuzzy domain division.Based on the above ideas,this thesis does the following work:(1)A progressive modeling method is proposed for hierarchical fuzzy inference systems.The proposed method selects input variables of each sub-system through independence test and inertia sorting with the contingency tables constructed with subtraction clustering.The sub-system construction process will stop when the system precision reaches the threshold value.The method can reduce the number of rules while maintaining the system accuracy.In the subsequent modeling process of hierarchical fuzzy inference system,the domain division results given by the previous subtraction clustering is maintained,which avoids repeated clustering and reduces the computational complexity in modeling.The experimental results show that the proposed method performs well and has universal applicability.(2)A method of progressive modeling hierarchical fuzzy inference system based on variable importance is proposed.The random forest model will be used to calculate the importance score of variables,and the input variables of each sub-system will be selected by the importance score of variables.The sub-system construction process will stop when the system precision reaches the threshold value.This method can not only improve the accuracy of hierarchical fuzzy inference system,but also solve the problem of the ineluctability of random forest through hierarchical fuzzy inference system.Through experiments,the applicability of variable importance score in hierarchical fuzzy inference system is verified by comparing the importance score with the hierarchical fuzzy inference system using other variable ranking methods.The proposed algorithm is compared with other existing modeling methods,and the comparison results prove the feasibility and strong performance of the proposed algorithm.