Using k-harmonic means clustering for the initialization of the clustering method based on one-class support vector machines

The initialization of one clustering method based on one-class support vector machines often employs random samples. This way can lead to the unstable clustering results. In this paper, the k-harmonic means clustering takes the place of this random initialization. To investigate the effectiveness of the novel proposed approach, several experiments are done on one artificial dataset and two real datasets. Experimental results show that our presented method can not only obtain the stable clustering accuracies, but aloes improve the clustering performance significantly compared to other different initialization, such as random initialization and k-means initialization.