Fuzzy Clustering with Fitness Predator Optimizer for Multivariate Data Problems Shiqin Yang and Yuji Sato Graduate School of Computer and Information Sciences Hosei University, Tokyo, Japan Abstract. Fuzzy c-means (FCM) is the most common fuzzy clustering model and uses an objective function to measure the desirability of partitions. However, if the data sets contain several noise points, or if the data sets are very high di- mensional, the iteration process of optimization the FCM model often falls into local optima solution. To avoid this problem, this paper proposes a new hybrid fuzzy clustering algorithm that incorporates the Fitness Predator Optimizer (FPO) into the FCM model. FPO is a new bionic-inspired algorithm to avoid premature convergence for the multimodal optimization problem. The excellent probability of finding the global optimum of FPO enhances the quality of fuzzy clustering. Five benchmark data sets from the UCI Machine Learning Repository are used to compare the performances of proposed FPO-FCM with FCM and a hybrid swarm algorithm based on Quantum-behaved PSO. Experimental results show that the proposed approach could demonstrate the desirable performance and avoid the minimum local value of objective function for multivariate data type clustering problems. Keywords: Fitness Predator Optimizer, Fuzzy C-means Model, multimodal op- timization problem, hybrid fuzzy clustering algorithm 1 Introduction The most common popular data mining techniques discussed are clustering and classi- fication. The clustering aims at identifying and extracting significant groups in under- lying data, which is an unsupervised learning method. In the field of clustering, Fuzzy c-means (FCM) is one of the most popular algorithms. Although FCM is extensively used in literature, it suffers from several drawbacks. The objective function of the FCM is the multimodal function which means that it may contain many local minima. Con- sequently, while minimizing the objective function, there is possibility of getting stuck at local minima or saddle points. In addition, the performance of the FCM depends on the initial selection of the cluster center. To increase the probability of finding the global optimum, various alternative meth- ods for the optimization of clustering models were suggested in the literature. Some researchers adopt the stochastic methods such as evolutionary or swarm-based methods to increase the global convergence ability of fuzzy clustering. In [7], authors used a Fuzzy c-means algorithm based on Picard iteration and PSO (PPSO-FCM) to improve the performance of FCM. In [8], a hybrid data clustering algorithm based on PSO and