Handling of Missing Values in FCM Clustering-based ANFIS with Partial Distance Strategy
Katsuhiro Honda, Satoshi Hyakutake, Seiki Ubukata, Akira Notsu · 2022 Joint 12th International Conference on Soft Computing and Intelligent Systems and 23rd International Symposium on Advanced Intelligent Systems (SCIS&ISIS) · 2022
Adaptive Network-based Fuzzy Inference System (ANFIS) is a promising model of explainable neural networks, which constructs Takagi-Sugeno fuzzy inference system based on neuro learning, but cannot work well with incomplete datasets including missing values. In this research, a novel approach of making ANFIS robust against missing values is proposed by modifying the premise part with Fuzzy c-Means (FCM) utilizing the partial distance strategy. By calculating the premise fuzzy memberships of incomplete objects having missing values with FCM criterion using partial distances, the proposed ANFIS variant can produce ANFIS outputs even if some objects include missing values in their input observation. The characteristics of the proposed model are demonstrated through a numerical experiment with a time series data.