Extraction of Fuzzy Rules from GIS Data using Fuzzy Nervous Network for Identifying Decayed Tissues in Tehran's 5th and 11th Districts

In this research, a method for extracting effective and interpretable fuzzy rules from GIS data using a neuro-fuzzy system is presented. The fuzzy model has passed through three stages to achieve high accuracy and interpretability. In the first stage, the primary weights of the neuro-fuzzy network were obtained using the FCM clustering algorithm. In order to categorize the educational data in the second phase, a neuro-fuzzy CANFIS system was used and genetic algorithms were utilized to overcome the fuzzy models loss of interpretability. The proposed method has been tested on the data of 5th and 11th districts of Tehran for the diagnosis of decayed tissues. The issue at hand is of the type of classification and the aim is to determine the degrees of membership of the textures in each of the classes. The decay of tissues has been examined in 4 categories including low, moderate, high and very high decay. A total of 300 educational samples were used, and after network training all data were categorized correctly and with RMS = 0.0045. The results show that the proposed method in this study has high accuracy and interpretability and is capable of generalization to issues in which sufficient knowledge of the target system is not available.