Twin Support Vector Machine Clustering Algorithm And Its Application

Clustering is one of the basic problems in the field of pattern recognition and machine learning and has been widely used in practical problems.The twin support vector clustering(TWSVC)was proposed in 2015 and has been concerned and studied by many scholars.It has gradually become a research hotspot of clustering algorithm.However,the clustering effect of TWSVC is greatly affected by noise or outliers,which often has poor robustness.In this paper,TWSVC is studied from the perspective of constructing weight function and loss function,and a new robust TWSVC algorithm is proposed,which can effectively improve the clustering accuracy of the algorithm and reduce the influence of noise or outliers on model performance.The specific research contents of this paper are as follows:1.In order to reduce the influence of outliers on clustering effect,a new weight function is constructed to give different punishments to other sample points.Improved least squares twin support vector clustering(ILSTWSVC)is proposed.The algorithm only needs to solve a series of linear equations without solving quadratic programming problems.Experimental results show the effectiveness of ILSTWSVC algorithm.2.RampTWSVC algorithm is an improved TWSVC algorithm based on the Ramploss function.However,the Ramploss function is symmetric and the same penalty is applied to the data points on both sides of the clustering center plane,so the clustering performance is poor on the data sets with noise.In order to solve this problem,an asymmetric Ramploss function is constructed in this paper,which inherits the bounded property of the Ramploss function.In this way,the influence of data points far from the cluster center plane on the cluster plane can be reduced effectively,and different penalties are applied to data points at different locations.The improved ramp-based twin support vector clustering(IRampTWSVC)algorithm is proposed.The algorithm is solved by alternate iterative algorithm,and the experimental results verify the effectiveness of the proposed algorithm.3.In order to further improve the performance of RampTWSVC algorithm,an asymmetric distribution loss function is constructed based on the asymmetric Ramploss function,which can capture the distribution information of data while applying different punishments to data points.Based on this,a plane-based clustering with asymmetric distribution loss(ADPC)algorithm is proposed.Proximal differential-of-convex algorithm(pDCA)is used to solve the algorithm,and experimental results verify the effectiveness of the algorithm.