A Monte Carlo manifold spectral clustering algorithm based on emotional preference and migratory behavior

Mingzhi Dai, Xiang Feng, Huiqun Yu, Weibin Guo, Xiuquan Li

Abstract: Inspired by various behaviors of creatures in nature, numerous efficient bionic algorithms are designed for dealing with complex clustering problems. As a population-based intelligence bionic optimization model mixes individual information interaction and physical mechanism, the emotional preference and migration behavior clustering (EPMC) algorithm had proposed for dealing with clustering tasks. It is superior to multiple classic clustering algorithms and obtains epoch-making clustering effects. However, EPMC still shows premature convergence and ineffectiveness in balancing exploration and exploitation. To further enhance the searchability and the performance during clustering, we proposed a Monte Carlo spectral clustering algorithm for emotional preference and migratory behavior optimization named MCSC-EPMC. Specifically, we first incorporate the spectral clustering strategy based on Laplacian eigenmaps to assist in updating the individual. Second, a Monte Carlo statistical data theory is utilized to simulate the cluster center point and help to approach the optimal. In addition, the theoretical analysis and convergence property of MCSC-EPMC are discussed. Numerous experiments were performed to compare the proposed MCSC-EPMC with the other seven clustering algorithms on several standard datasets. And the experimental results on several standards show the effectiveness and feasibility of MCSC-EPMC. It also enhanced the clustering performance by about 6.036% compared to the EPMC.