The comparison of classification performances of hierarchical and non-hierarchical fuzzy models

In this study, it is aimed to compare classification performances of hierarchical and non-hierarchical fuzzy models by using different membership functions on simulation and real data sets. In the application step of the study, algorithms are created by using MATLAB software for simulation practice. Classification performances of the hierarchical and non-hierarchical fuzzy models created using both data sets generated by means of these algorithms and real data set are compared according to sensitivity, specificity, accuracy and root mean square error (RMSE) criteria. In the end of simulation practices, it is determined that sensitivity, specificity and accuracy of the non-hierarchical models are higher; RMSE of the non-hierarchical models are lower than hierarchical fuzzy models. However, it is found that the rule base of hierarchical fuzzy models are larger than that in non-hierarchical models. Also, it is determined that the membership functions that give the best result are triangular and bell. In the real data set application - similar to simulation results - it is found that the specificity, accuracy of the non-hierarchical models built by different membership functions are higher, RMSE of the non-hierarchical models are lower than the hierarchical models. It is stated that models that give the best classification results are built by Gauss or bell membership functions. Consequently; non-hierarchical fuzzy logic technique give better classification performance than hierarchical fuzzy logic technique. Consequently; non-hierarchical fuzzy models classify better than hierarchical fuzzy models. Since structure of the data, distribution of variables and correlation between them, how to seperate variables into sub-categories in fuzzification step and which method to use are to be studied on in detail, experienced expert help should be taken. Keywords: Hierarchical Fuzzy Model, Non-Hierarchical Fuzzy Model, Classification, Simulation.